Joint matrix constraints barely change the prediction ranges
All 48 sectors pass correctness and independent endpoint checks. Complete widths improve by less than 0.1%; four of six groups still pass the unchanged target. Both specified-field globe groups still miss it. The additional relationships are mathematically useful checks, but they do not solve the remaining width.
These are tests of the uncertainty calculation. Passing a width target is not evidence for that physical candidate. The completed 53 aircraft comparisons remain unchanged: all favor globe geometry under their stated assumptions.
The study finished in 156.1 seconds. All 288 LP reservations and 1,152 exact support reservations reconcile with sealed events. No new native prediction, integration bound, physical oracle, original-aircraft read or observed fit ran. Every original coordinate, coefficient and remainder allowance remains.
Further variations on linear matrix cuts are a lower priority. Next, inspect whether the extreme combinations admitted by the calculation can actually come from one physical calibration. A disagreement would identify missing dependence between error terms; agreement would show that the remaining range cannot be removed merely by tying those terms together. This diagnostic will not change the acceptance target or label old failures as successes.