This paper introduces a denominator-recovery audit framework for high-correlation claims in a GHZ three-body model, aimed at separating mechanism gain from post-selection bookkeeping gain. Within a unified pipeline, we compare `none` and `energyweighted` denominator modes and run a two-stage search (90 deg coarse + 2 deg fine), while jointly auditing `F` and `coincidenceᵣate` trade-offs. Under the current configuration (`samples=80000`, `numbacpu`, fixed gating), the best fine-stage candidate reaches only `F=0. 085396`, remaining far from the target `F=4` with `|err|=3. 914604`. Target-attainment decomposition consistently indicates correlation-shape mismatch as the dominant bottleneck rather than coincidence sparsity. Robustness indicators (bootstrap and seed sweep) do not show convergence toward 4 either. We therefore report a method-level conclusion: within the tested model family and protocol, no evidence supports an `F->4` transition. Strong GHZ-style claims should be evaluated through joint auditing of denominator logic, sample-inclusion rules, and retained failure cases. The contribution is a reproducible, ledger-style audit workflow for GHZ correlation studies. Code repository: https: //github. com/tomnattle/chain-explosion-model
hui wang (Sat,) studied this question.