CodeBLEU scored 91% for both approaches while one hallucinated APIs 56.4% of the time.

For monolith-to-microservice translation, a graph pipeline built on tree-sitter ASTs, a Spanner property graph and a Gemini context cache dropped API hallucination from 56.4% to 16.2% and raised dependency resolution from 34.8% to 65.9%. Text-overlap metrics rated both at 91%, hiding structural failure behind syntactically plausible broken code. The graph approach isn't free: dense global context pushed the model into defensive over-engineering, cutting cyclomatic complexity consistency from 71.6% to 46.7%.