How often have your RAG issues actually turned out to be document parsing issues?

I’ve been thinking about this a lot lately.

When a RAG system gives bad answers, the first instinct is usually to look at chunking, embeddings, retrieval, or the model.

But sometimes the problem started earlier.

If the parser already destroyed the table structure, heading hierarchy, or reading order, retrieval is working with bad input from the beginning.

Curious how often others have run into this.

Was the real bottleneck actually the ingestion/parsing layer?