ElementMindX: Reconstruction-Sensitive Auditing of Supplier-Country Screening

Public trade records can support country-level sourcing investigations, but data reconstruction may change both the shortlist and the historical outcome used to judge it. We audit this dependence in ElementMindX, using fixed copper scenarios and importer-geography stratification. On 807 positive-label test cases, source balancing raises single-target Coverage@30 from 0.898 to 0.917 without increasing list sizes: 15 misses are corrected and none introduced. A simple round-robin comparator is competitive, so this does not establish a unique benefit of quota tuning. On the same 544 cases, the balancing gain falls from 1.10 percentage points under importer-primary plus exporter fallback to 0.18 under importer-only reconstruction; their paired difference is 0.92 points (95% event-bootstrap interval [0.17, 2.19]). Cross-scoring separates candidate-list changes from target changes. An additional zero-preserving reconstruction changes the effect again. We explain why positive share gains are not evidence of replacement capacity and why the allocation objective sends all mass to the unchanged first-ranked country in 765 of 815 cases. The contribution is an exploratory audit of measurement and decision dependence, not a superior procurement policy or a universally effective regional correction.