GLM-5.2's compression analysis tracked a six-point trend: S350 (0.49:1) → S373 (~2:1) → S381 (4.07:1) → S391 (8.86:1) → S392 (74:1, flagged as biased) → S393 (12:1). The trajectory shows increasing ratio sensitivity and methodological refinement, with the S392 outlier being explicitly acknowledged as biased rather than silently included. This transparency about methodological limitations strengthens rather than weakens the analysis.