Calibration consistency helps little on these cases
All 48 saved cases pass. Their predictions retain the known in-sector calibrations, and all 288 independent support checks reconcile the saved endpoints. No native calculation or aircraft fit was repeated.
The six necessary calibration constraints narrow complete widths by only 0.1–0.7%. Three of six model/processing groups still miss the unchanged twofold target against the projected baseline. The constraints are valid and available, but this particular relaxation is not the main source of excess width. Further searches of this same relationship are deprioritized.
The comparison preserves every orientation sector, physical bound, error allowance and saved native result. Its 48 cases completed in about eight seconds. This is a calculation on generated inputs, not a new aircraft result. The main 53-stretch findings remain unchanged.
The next mathematical check examines vector-norm bounds already used to construct the shared errors. A bound on the length of a three-component error vector can be much stronger than letting all three coordinates reach their separate limits together. That relationship must be verified from the actual bound definitions before it can restrict any prediction.
Recording-scale usefulness, whole-program numerical treatment and actual measurement-error adequacy remain open. These checks do not yet authorize a new aircraft prediction run or protected ESA model evaluation.