Localized Residual Architecture in B → K* μ+μ− Angular Observables: A Data-First Diagnostic Audit of Public HEPData Records

This technical note presents a data-first diagnostic audit of public HEPData records on angular observables in B → K* μ+μ− decays. The aim is not to propose new physics, reinterpret the underlying flavour dynamics, or replace global-fit analyses. Instead, the note examines how measurement–theory residuals are distributed across observable–q² cells when multiple theory baselines are available. The main case is the CMS 2025 public HEPData record associated with the angular analysis of B0 → K*(892)0 μ+μ− in proton-proton collisions at √s = 13 TeV. The LHCb 2021 B+ → K*+ μ+μ− record is used only as a supporting morphology benchmark for optimized angular observables. The audit classifies residual nodes according to local pull magnitude, theory-baseline availability, cross-model persistence, and sign agreement. The results suggest that the residual structure is neither globally distributed across all angular observables nor reducible to a single P′5-centered anomaly. Under the adopted diagnostic criteria, P′8 in the q² = 2.0–4.3 bin emerges as the cleanest cross-model residual node in the CMS 2025 public record, while P′5 remains central but not exclusive. This finding should not be read as identifying a new P′8 anomaly. It only indicates that, within the public CMS HEPData rows and the diagnostic criteria used here, the P′8 | q² = 2.0–4.3 cell has the strongest residual-node profile. The contribution is methodological and reproducible, based entirely on public HEPData records. The accompanying reproducibility bundle contains code, notebook, CSV outputs, figures, HEPData JSON/YAML files, and traceability tables.