Observable-Class Derivation Mismatch and the Heisenberg Transfer No-Go: From Proxy Capacity to Block-Level Span Dynamics

The O3–O7 derivation chain proposes, on a changing-degree LPS family, the relation β* = 1/(δ + 1/2) linking the capacity decay exponent δ to the cascade growth exponent β* governing charged-lepton mass ratios. O13 established that the exact Weil-block capacity exponent δ̂exact(q) ∈ [3.6, 4.8] does not drift toward the target window δ ∈ [7.4, 10.6] as q grows, ruling out a finite-size explanation of the tension; O12's independent corrective revision derives, in closed form, the intrinsic asymptotic value the finite-q fits are converging toward, δexact=3, confirming this analytically rather than by extrapolation, and shows the fitted values carry no central-coordinate dependence of any kind. The present paper identifies a distinct, independent structural reason — unrelated to the central coordinate — that the tension cannot be resolved by refining the block-level measurement, and proves a non-transferability result: the O6 growth equation dp/dn Rₙ \, cBI √p(n) derives β* from a scalar global redundancy functional Rₙ, whereas the exact observable measured in O12–O13 is a block-averaged capacity over a finite sampled subset of generic-triple Weil representations, each of internal dimension q. We prove a non-equivalence theorem establishing that the block-averaged exponent δ̂exact does not, in general, coincide with the scalar exponent α entering the growth equation. We define an effective block-weighted redundancy Rₙᵉᶠᶠ, prove it satisfies the exact identity Rₙᵉᶠᶠ=Σ̄ₙ·(1+Vₙ) for the natural dynamical weights (where Vₙ is the inter-block variance ratio already reported in O12/O13), and show this bounds the aggregation exponent by δ̂exact whenever log(1+Vₙ) has non-negative OLS slope against log n over the fitting window (in particular whenever 1+Vₙ is non-decreasing there), a condition checkable from existing pipeline data, so this specific, motivated reweighting is constrained, subject to that check, from repairing the gap by aggregation alone. The native identity is instead Δ rₙ=|Sₙ|Σₙ Σₙ nD-1. Neither the Born–Infeld factor cBI√p nor the changing LPS valence has a native carrier on the fixed-degree Heisenberg cascade. Moreover, a constant factor q-η cannot alter an intra-q slope, and direct exponent matching in the transplanted equation does not yield the legacy reciprocal formula. The former C1/C2 outputs are therefore retained only as phenomenological diagnostics. Block-level redundancy measurement remains useful for the aggregation question, not as a test of a native capacity-to-rate transfer.