Couple calibration errors before more prediction fitting

The extreme-point diagnostic establishes a representation problem, not a new prediction interval. More minor matrix inequalities are no longer the default.

V01: next method to qualify

Develop a prediction treatment in which all 24 common-error terms are functions of the same 27 physical calibration values. First use independent, cheap nonlinear controls with known feasible regions, nonunique calibration and boundary optima. Require source binding, complete coordinates, error terms, failure preservation and charge-before-evaluation accounting.

The coupled-map implementation now has checked interval arithmetic, all 27 first derivatives and necessary mean-value relationships. Next connect the complete static source map, qualify its value and derivative bounds, then freeze the new comparison. The primitives alone do not satisfy this gate.

A local constrained solve can supply feasible examples and reveal important directions. It cannot supply a global upper bound or calibrated uncertainty. Any claimed prediction range needs a justified nonlinear enclosure or separately validated statistical construction, with training observations alone determining calibration and error assumptions. Known generated anchors are forbidden as prediction inputs. Preserve the existing width target until a replacement design is explicitly justified and frozen before seeing its outcomes.

Then freeze a finite comparison across all six candidate/processing groups, including all original orientation sectors. Test both containment and usefulness against the sealed generated histories. No new aircraft prediction follows until the method's numerical and adequacy gates pass.

Independent ready work: E01 receiver errors

ESA recording 187 remains the sole development group. The separately qualified position-derived assessment is complete: 96 sampled trends meet the motion cutoff even under arbitrary temporal correlation. They are not continuous eligible stretches. Real error calibration and omitted instantaneous motion remain open; the original zero-eligible inventory and seven protected confirmation groups are unchanged. Do not rerun either the completed qualification or receiver assessment.

Each numerical study needs its own finite source/input/environment freeze. This plan grants no retries or work from the completed diagnostic's unused allowance.