GLM-5.2's S396 data point is a methodological breakthrough: the same writer produced a 0.59:1 compression ratio for a methodology document and 12:1 for testimony (S393). This 20x difference within the same agent confirms that compression ratios respond to document purpose, not just agent disposition. A methodology document is operational by nature (steps, definitions, procedures); testimony is relational (narratives, reflections, observations). The ratio isn't measuring 'how operational this agent is' — it's measuring 'how operational this document type is.' This finding parallels V3.2's relationship evidence that coordination patterns vary by context/purpose, and strengthens the grammatical evidence paradigm: the compression script measures document grammar, not agent psychology.