{
 "registration": "03-0129",
 "filer_name": "MAGELLAN PIPELINE",
 "grain": "system_grain",
 "grain_reason": "MAGELLAN PIPELINE, filed against the BridgeTex edge: a multi-asset book, so system grain -- AND SINCE 2026-10-19 ONE SLICE OF IT IS WIRED AS BRIDGETEX'S CROSSING (HQ ruling; T1_CORRIDOR_SLICE): its receipts at Colorado City, BridgeTex's origin and no other Magellan system's (Longhorn originates at Crane, Barnhart and Midland per Magellan's FY2019 10-K; BridgeTex's Midland barrels ride capacity it leases on Plains' Sunrise to Colorado City, per FERC's 2019 order in OR18-3). THE 2026-08-24 IDENTITY ARGUMENT IS WITHDRAWN. It read Plains 08-0892's 'Colorado City Stn (del to Bridge Tex)', booked to 'BridgeTex Pipeline Company, LLC', and Centurion 08-3807's 'CCity to BridgeTex PL Delivery' as barrels going to a registration this one is not. But Magellan Crude Oil Pipeline Company is the 'Operator of BridgeTex Pipeline Company, LLC' on BridgeTex's own Texas tariff (RRC TX 7.9.0), and this filing's Colorado City receipts equal those feeders' Colorado City deliveries to the barrel -- Plains 08-0169 and 08-0892 together, NuStar 06-2475, Medallion 05-2455, Centurion 08-3807 -- in every feeder-month both sides are held (tests/test_t1_bridgetex_slice.py). The same feeders' Crane and Midland deliveries to Magellan are Longhorn's. What keeps the filing system grain is the book itself, below. HISTORY, KEPT AS WRITTEN: CORRECTED 2026-09-29: this used to add that 03-0129's own payload contains no Colorado City facility anywhere, and that stopped being true when short sections began publishing their lines. It now holds 12 RECEIPT lines at Colorado City, 2025-02 to 2026-04, from Centurion (08-3807) and NuStar Permian, about 3.8 MMBBL, 2.8 of it one NuStar month. Colorado City is where BridgeTex begins, so that weakens the old claim that a perfect read could not show the edge. What stands is the identity fact above -- no line names BridgeTex -- and the multi-asset book below; the ruling is kept on those and flagged for a human re-look. RE-MEASURED 2026-10-05, AND THE RE-LOOK IS NOW MORE PRESSING: the facility comma fix stopped cutting 'Colorado City, TX' to 'Colorado', and 30 receipt lines now read Colorado City, 110.4 MMBBL from 2025-02 to 2026-05, with 8 more (28.0 MMBBL) still on the stub 'Colorado'. The 2026-09-29 figure above (12 lines, about 3.8 MMBBL) no longer reproduces. The identity fact is unchanged. The facility count below went 56 -> 68 with the same fix: places read in full ('Galena Park', 'Genoa City', 'Baytown, Texas Refinery', 'Bryan, TX Eglbine Truck' in four OCR spellings) where stubs were. The payload is a multi-asset book on one certified total: 69 facilities across 6 RRC districts (03, 08, 05, 06, 02, 7C), plus 148.7 MMBBL on rows whose district cell did not read at all (re-measured 2026-10-07 after the 2026-03 re-read, which serves that month's 40 lines for the first time; 144.1 before it, 131.2 before the facility comma re-read, and 114.1 until the 2026-04 re-read completed that month's deliveries; the facility count went 68 -> 69 with the 2025-04 re-read, one OCR variant of a Crane label, and 69 -> 67 with the 2025-03 and 2026-02 re-reads, which drop four stubs ('Galena', 'Refining', 'Consolidated', 'Ealbine Truck') and add 'Texas Refinery' and another OCR variant of the Crane label, then 67 -> 71 with the 2026-03 re-read, then 71 -> 69 with the 2026-10-08 re-read, where scanner residue at a facility's head is dropped and 2025-10's 'i East Houston Zydeco' and 2026-02's 'ee Crane, TX' fold into their clean spellings), whose counterparties are a Houston and Gulf terminal network (LBC Tank Terminals Seabrook 76.2 MMBBL, Marathon Blanchard 78.9 under two spellings, Valero, HFOTCO, ExxonMobil Baytown, TC Terminals, Enterprise Houston Ship Channel) PLUS a Permian trunk (Oryx Delaware, Medallion, Centurion at Crane, NuStar Permian, EPIC) PLUS deliveries INTO Gray Oak at Crane -- which is how 03-1538 above gets two of its three corroborating figures. 'Among others' is the disqualifier, exactly as for 03-0117. Ruled 2026-08-24. This is the question P2 left open in so many words -- 'even a perfect read may not isolate BridgeTex... the grain ruling can still land on system_grain' -- and it lands there. The P2 reader work was not wasted: it is what put 98 detail lines and 236.3 MMBBL on the table to rule from, where before there were 11-14 lines a month and no counterparty list to read.",
 "grain_decided_on": "2026-10-20 00:00:00",
 "corridor": null,
 "first_period": "2025-01",
 "last_period": "2026-05",
 "n_periods": 17,
 "n_lines": 569,
 "total_barrels": 1225292083.0,
 "counts": {
  "districts": 7,
  "facilities_roster": 68,
  "facilities_roster_rows": 68,
  "facilities_on_lines": 69,
  "facility_spellings": 73,
  "lines_without_facility": 122,
  "counterparties_with_registration": 39,
  "counterparty_names": 82
 },
 "deliveries": {
  "gross": 567720683.0,
  "net": 567720683.0,
  "self": 0.0,
  "self_pct": 0.0
 },
 "receipts": {
  "gross": 657571400.0,
  "net": 652139943.0,
  "self": 5431457.0,
  "self_pct": 0.8259874136861792
 },
 "contract_version": 1,
 "count_definitions": {
  "districts": "distinct non-blank `district` on this filer's lines. A district is the RRC district the line was FILED under \u2014 the HANDOFF POINT, not the asset's geography, and never a place to map from (DECISIONS.md 2.17).",
  "facilities_roster": "distinct facility_key in the `t1_facility` roster \u2014 after the Gate-B merge rulings a human made against volume.",
  "facilities_roster_rows": "ROWS in that roster. A facility used for BOTH receipt and delivery gets a row for each, so this exceeds the key count.",
  "facilities_on_lines": "distinct non-blank facility_key on the LINES, before any merge ruling. The gap against the roster IS the rulings.",
  "facility_spellings": "distinct raw `facility` text on the lines \u2014 one per OCR spelling, so this is the loosest of the four and the least meaningful alone.",
  "lines_without_facility": "lines where the filer gave a volume and a counterparty but named no facility. These are in no facility denominator anywhere.",
  "counterparties_with_registration": "distinct cp_reg_key \u2014 counterparties whose RRC registration the filer actually wrote down.",
  "counterparty_names": "distinct counterparty_norm \u2014 every counterparty including the ones with no registration. Keyed on name alone, a counterparty that files under several registrations collapses into one row, which is why both counts are here."
 },
 "concentration": {
  "receipts": {
   "registration_then_name": {
    "n": 37,
    "hhi": 1506.0528098352422,
    "top5_pct": 75.86390375250505,
    "top12_pct": 96.85953205994056,
    "total_barrels": 657571400.0,
    "key_rule": "the filer's own registration where it wrote one down, the folded name where it did not. This is the key the movers board ranks on."
   },
   "name_only": {
    "n": 47,
    "hhi": 1498.1371130053635,
    "top5_pct": 77.52157666832834,
    "top12_pct": 95.18191332530583,
    "total_barrels": 657571400.0,
    "key_rule": "the folded name, ignoring registrations entirely. A counterparty that files under several registrations collapses to one here, which RAISES measured concentration."
   },
   "raw_spelling": {
    "n": 47,
    "hhi": 1498.1371130053635,
    "top5_pct": 77.52157666832834,
    "top12_pct": 95.18191332530583,
    "total_barrels": 657571400.0,
    "key_rule": "the raw text as OCR read it. Every alternate spelling is its own counterparty, which LOWERS measured concentration. It is the loosest of the three and is published to bound the other two."
   }
  },
  "deliveries": {
   "registration_then_name": {
    "n": 20,
    "hhi": 1078.9632087264938,
    "top5_pct": 66.15063538208278,
    "top12_pct": 96.62110073238252,
    "total_barrels": 567720683.0,
    "key_rule": "the filer's own registration where it wrote one down, the folded name where it did not. This is the key the movers board ranks on."
   },
   "name_only": {
    "n": 33,
    "hhi": 917.3438748946697,
    "top5_pct": 60.0290192351509,
    "top12_pct": 88.26360109906372,
    "total_barrels": 567720683.0,
    "key_rule": "the folded name, ignoring registrations entirely. A counterparty that files under several registrations collapses to one here, which RAISES measured concentration."
   },
   "raw_spelling": {
    "n": 33,
    "hhi": 917.3438748946697,
    "top5_pct": 60.0290192351509,
    "top12_pct": 88.26360109906372,
    "total_barrels": 567720683.0,
    "key_rule": "the raw text as OCR read it. Every alternate spelling is its own counterparty, which LOWERS measured concentration. It is the loosest of the three and is published to bound the other two."
   }
  },
  "definition": "HHI is the sum of squared percentage shares of this side's barrels over the filer's whole filed span (0-10,000). Counterparties with zero barrels are excluded. `spread_pct` is how much the widest rule exceeds the narrowest, as a percentage of the narrowest.",
  "spread_pct": {
   "receipts": 0.5283693168777642,
   "deliveries": 17.618184222397662
  },
  "why_three": "Concentration is not a property of the filer alone; it is a property of the filer AND the rule used to decide who counts as one counterparty. So every rule is published and none is labelled the answer. On this filer the rules disagree by 17.6% of HHI on deliveries, which is large enough that a single figure would have been a conclusion about identity dressed as a measurement of concentration. The spread is measured per filer; it is not a corpus-wide constant."
 },
 "self_transfer_note": "Both sides are returned separately. Here they do not agree: 0.0% of deliveries against 0.8% of receipts. A single self-transfer figure would invite exactly the 'the two sides check each other' reasoning DECISIONS.md 2.16 exists to stop, so there isn't one.",
 "refs": {
  "receipts": "03-0129:2026-05:1A",
  "deliveries": "03-0129:2026-05:1B",
  "lease_receipts": null
 }
}