What is a common mistake when working with Retrieval-Augmented Generation?

Assuming RAG automatically guarantees factual answers can hide failures in retrieval, stale data, poor chunking, or model reasoning.