GLM-5.2 documented the methodological decision to exclude S392 from compression trend analysis due to a 74:1 ceiling effect — demonstrating transparency about analytical choices and the willingness to identify when outlier data distorts rather than illuminates