Analytic, photographic and inconsistent-input evidence remain distinct.
Corresponding rows, longitudinal spans and fitted scalar spline coefficients.
Actual selected engine event/interval/mesh counts and numerical boundary defect.
Maximum pixel residual of fitted offsets or their implied polygon supports.
Common measured row-centre interval, in millimetres.
Independently generative coefficients and edit/coverage metadata.
Every training source identity; camera charts remain in the observation artifact.
Selected unpivoted Householder diagonal ratio; not a certified condition.
Fit evidence kept outside compact support-coefficient generation authority. Pixel residuals include the implied intersection's support, not only each independently fitted offset. Numerical event and mesh counts qualify the selected representation; they do not observe hidden physical topology.