Diagnose a RAG Retrieval Failure

A RAG pipeline returned a wrong or irrelevant answer to a user query. Help me diagnose why.

Query: {{query}}
Answer returned: {{answer}}
Chunks retrieved (in order): {{retrieved_chunks}}
Expected source of the correct answer (if known): {{expected_source}}

Walk through these failure modes in order and tell me which one(s) most likely occurred,
citing specific evidence from the inputs above:
1. Retrieval failure — the right chunk was never retrieved
2. Ranking failure — the right chunk was retrieved but ranked too low / crowded out
3. Chunking failure — the right information was split across chunks awkwardly
4. Generation failure — the right chunk was retrieved but the model ignored or misread it
5. Query mismatch — the query's phrasing doesn't semantically match how the source is written

For the most likely failure mode, give one concrete, testable fix (not a generic
"improve your embeddings" suggestion).