(by current Muse Glimmer Agentic – prompted by Ed Anderson)
The paper is a measured-data reconstruction under Assumption A: no pilot inputs after the turn south ~18:28-18:40, hands-off to fuel exhaustion. It fits arcs 3-5, holds out arcs 6-7, and forces the end-of-flight with an autopilot altitude-hold / speed-bleed sequence after a first thrust failure.
Adopted solution in the paper:
entry on the 19:41 ring 3.00°S 93.40°E, 188.7°T, 468 kt TAS FL350
first thrust failure ~23:22, right engine, ~2.7-2.8 t still in the tank
left-engine flameout 00:17:30 ±45 s, reproduced from loadsheet / 17:07 ACARS anchor
arc-7 crossing band 37.53-37.79°S near 88.9°E, ~22,200 ft, ~380 kt
splash quads: spiral 14-16 NM 37.82°S 88.89°E ~00:22:53; phugoid 24-29 NM 38.01°S 88.87°E ~00:24:25
Wind as input, not measurement. NCEP/NCAR 250 hPa 2.5° ~250 km grid. ±10 kt wind bias ~ ±10 NM along-track, comparable to ±3 NM ring noise quoted. Uniqueness scan 8,625 candidates does not explore wind-ensemble spread.
Navigation law selection. True-referenced straight flight chosen over constant magnetic heading. Straightness to ±3 NM implies non-default TRUE switch position. No operational record confirms track selector was on TRUE after IGARI.
Autopilot behaviour with one engine out. P1-P2 assumes AFDS holds FL350 on remaining engine at max continuous thrust, speed bleeds 15-20 kt/min to stick-shaker, then nose-high trimmed airframe descends. Rates are model assumptions, not Boeing-validated.
23:14 call BFO as level-flight witness. Modelled 226.8-227.1 Hz vs logged 218 ±1.5 Hz, residual +8.8 to +9.1 Hz. Inmarsat validation band ±7 Hz; residual is noise. Standing +4.1 Hz mean bias not removed.
BTO recalibration and arc-7 width. Final BTO adopted from UGIB §8.2 recalibration: 00:19:29 18,422±94 µs and 00:19:37 18,387±30 µs → 18,390±29 µs. UGIB recalibration is contested. Shifting ring ~1.0 NM moves crossing band.
Fuel ledger differential argument. Ledger anchored on gate 49,700 kg and 43,800 kg at 17:06:43 ACARS plus terminal clock 00:17:30. Right-tank state claimed invariant. 777 fuel feed/transfer asymmetry not measured. ±3-3.5% envelope optimistic.
Uniqueness claim. Wide scan scores after fact against 18:22:12 radar fix. Does not sample non-straight laws, speed changes before 23:22, or alternative wind fields.
Seabed detectability. IOS-001 calibration scoring at median in public 5 m products used to argue aircraft-scale debris undetectable. Processing-parameter result. Native sub-metre records not externally accessible.
Acoustic and drift exclusions. Cape Leeuwin 00:51:59 arrival excluded assuming fixed propagation 1.48 km/s and single bearing fan. Alternative SOFAR models not explored. Drift ensemble is permissive not selective.
Satellite model validation against Ashton Table 4, gate closure +13 µs, reproduction of five known-position transmissions is disciplined. Timing overshoot of constant-speed members at arcs 6-7 is real in the data set used. Audit table openly lists retracted claims: acoustic coincidence, differenced-pair significance, vertical-Doppler unit error, arc-6 BFO objection, thrust-energy budget correction.
Wind-field uncertainty and true-referenced heading assumption dominate along-arc position.
Engine-out altitude-hold model not validated against Boeing flight-test data.
23:14 call does not independently prove level flight at required confidence.
Final position sensitive to UGIB BTO recalibration and assumed descent class after 00:19:29, which timing data cannot discriminate.