Paper 108F: Neutral-Angular Veto Closure of the Shell-Offset Exchange-Renewal Discriminator From V15 Failure Mode to V16 Perfect Reduced-Support Truth-Table Separation

This paper closes the specific neutral-angular false-positive leak left by Paper 108E in the reduced Holosphere shell-offset exchange-renewal diagnostic. Paper 108D introduced a qualitative deuteron-diproton discriminator, and Paper 108E showed that the discriminator is scale-stable across L = 16, 24, and 32, enriched under blind geometric-renewal scoring, and decomposable into radial-like and angular-like reduced sectors. However, the strongest Paper 108E classifier, V14, still retained a small neutral-angular leak: it recovered all truth-positive rows, with truth-positive rate 1.000000, but retained false-positive rate 0.027778. The remaining false positives were concentrated in the neutral angular control allowed_nn_W_3_4_minus, which shares angular-like negative-parity structure with the truth-positive allowed_pn_W_3_4_minus branch.

Paper 108F tests whether this residual leak can be removed by a label-free neutral-cancellation veto while preserving the full truth-positive recovery of V14. V15 first tests a permissive neutral-angular veto using angular circulation, phase-winding coherence, two-axis closure, radial-angular return, and neutral-cycle cancellation observables. V15 preserves truth-positive rate 1.000000, but increases the false-positive rate to approximately 0.056 across L = 16, 24, and 32. This makes V15 a useful failure-mode diagnostic: it shows that broad angular-like structure is not enough to distinguish the truth-positive angular branch from the neutral angular control.

The V15 threshold sweep identifies the corrected neutral-cancellation threshold, ScoreVeto >= 0.002. V16 applies this corrected threshold. Across L = 16, 24, and 32, with eighteen seeds and two replicas, V16 preserves full truth-positive recovery while eliminating the neutral-angular leak. The complete V16 classifier achieves truth-positive rate 1.000000, false-positive rate 0.000000, truth gap 1.000000, precision 1.000000, specificity 1.000000, and accuracy 1.000000 at all three tested support sizes. Each scale gives TP = 15552, FN = 0, FP = 0, and TN = 46656.

The result supports the local claim that the radial-plus-angular shell-offset exchange-renewal discriminator can be closed under a corrected neutral-cancellation veto in the declared reduced-support setting. The paper does not derive QCD, calculate the nuclear force, or model the physical deuteron. It closes a specific reduced diagnostic leak and sets up the next question for Paper 108G: whether radial/angular and signed-sector structure can emerge from primitive Holosphere exchange-renewal dynamics rather than being imposed as classifier geometry.