This paper tests whether the Paper 108D shell-offset exchange-renewal deuteron-diproton discriminator survives stronger robustness pressure. Paper 108D showed that a reduced Holosphere shell diagnostic could distinguish proton-neutron-like and composite proton-neutron-support channels from same-positive, neutral-only, broken-mirror, and off-ladder controls under a qualitative truth-table rule. Paper 108E asks whether that result remains visible when the support size is increased, when scoring is performed before truth labels are revealed, and when the truth-positive class is decomposed into radial-like and angular-like exchange sectors. The audit proceeds through four stages. V11 computes a blind geometric-renewal score across L = 16, 24, and 32. It recovers all truth-positive rows at every scale, with truth-positive rate 1.000000, but admits nonzero false positives from neutral-only and broken-mirror controls, with false-positive rates around 0.235 to 0.240. The top 1%, 5%, and 10% of blind-score rows remain entirely truth-positive across all tested scales. V12 adds a signed spin-isospin proxy. This removes all false positives at the default threshold, with false-positive rate 0.000000, but preserves only two-thirds of the truth-positive class, with truth-positive rate 0.666667. V13 shows that the missing one-third is not random loss, but an angular-like branch. V14 adds an angular-role score and recovers all truth-positive rows, giving truth-positive rate 1.000000, false-positive rate 0.027778, truth gap 0.972222, precision 0.923077, specificity 0.972222, and accuracy 0.979167 across L = 16, 24, and 32. The result supports a layered interpretation of the reduced shell-offset exchange-renewal mechanism. Blind geometry and renewal identify candidate finite-distance bonds but are too permissive by themselves. Signed spin-isospin structure rejects broad false positives but selects only the radial-compatible two-thirds of the truth-positive class. Radial-angular decomposition explains the missing branch, and angular-role cleanup recovers it with a small neutral-control leak. The paper therefore supports the claim that the Paper 108D discriminator is not merely an L = 16 label-conditioned artifact. It is a scale-stable reduced-support structure whose full expression requires both radial/spin-isospin and angular-role sectors. The result is not presented as a derivation of QCD, the physical nuclear force, or the deuteron binding energy; it is a robustness and sector-decomposition audit within the Holosphere shell-support program.