By now, RAG is an accepted and well established standard for addressing data relevance for in-context learning. But there are concerns regarding model behaviour when inaccurate data is retrieved.