Gravitational Manipulation Theory v2.1: Locating the Provenance Ceiling in Apparatus-Native Torsion-Pendulum Reconstruction

Gravitational Manipulation Theory (GMT) v2.1 investigates whether the apparatus-native forward operator of the HUST-2020 torsion-pendulum experiment can be reconstructed from the public record. GMT v2.0 reconstructed the native field and force response of screened-scalar mechanisms but found that native exclusion reconstruction stalled at the apparatus-forward link. GMT v2.1 therefore treats this provenance ceiling itself as the research object. The primary record of Tan et al. (2020) is audited into confirmed, estimable, and unidentified geometric degrees of freedom. A finite-geometry forward operator mapping an input force to the 8ωd torque harmonic is reconstructed as far as the public information permits and tested against three pre-registered validation gates: model completeness, the published aligned Newtonian null, and the published non-null response at 250 and 500 μm. Across a pre-registered expanded public prior of 56 geometries, 0/56 pass the joint gate. Two coarse-grid near-null candidates are rejected under targeted grid refinement; their non-null responses remain outside the validation bands by 8.5–27.7×. The resulting verdict is: Ceiling Retained at the geometry / operator-definition layer. The public record does not uniquely specify enough finite geometry to define a validated null-cancelling apparatus operator. This locates the provenance ceiling one layer upstream of the exclusion-propagation ceiling identified in GMT v2.0. A separate representative-base counterfactual study scans three fine alignment parameters over 135 points and yields 0/135 joint closure. The best-tuned null remains approximately 27× above the adopted null cap. Within this representative-base scope, precision on these three known knobs alone does not close the gap; the missing information instead points toward the compensation-mass topology and absolute apparatus placement. The accompanying supplement contains the finite-geometry operator and gate code, primary digitizations, pre-registration and provenance records, the complete 56-member sweep, targeted convergence audit, counterfactual results, and Figures 1–6. The work distinguishes explicitly between primary, reconstructed, and counterfactual quantities. No experimentally demonstrated gravitational manipulation or engineering controllability is claimed.