Separate SIHUMIx BP5 hypothesis from measured community output #1769

SIHUMIx_Human_Intestinal_Model_Community.yaml (CommunityMech:000080; PMID33622394 and doi:10.3390/microorganisms7120641): genomic adjacency does not show BP5 causes neighboring enzyme activity. The in silico knockout removes rxn15467, not BP5; the authors explicitly leave BP5 mediation hypothetical. Remove expression-to-model and model-to-observed-SCFA arrows while retaining the three bounded observations. Remove COLONIZATION_FACILITATION from expression and CROSS_FEEDING from bulk SCFA measurements; retain explicitly computational exchange predictions without claiming in-vitro transfer. Account for differing single/community growth conditions, positive but limited synthetic-CR2 assay evidence, and the distinct transit-time experiment. No new topology.

Dedup: fresh all-state issue titles/bodies and ignored/hidden-inclusive local search with .git/docs/site/references_cache excluded; relevant caches separately inspected. Matching issues read, not duplicates of these findings. Comments/PRs not exhaustive. Source-based self-adversarial review, not independent approval. Graph-repair issues close only after protected-main merge.