And many times, after an error is pointed out and an LLM offers an "explanation" for what happened, the LLM then gives the exact same erroneous result.

The LLM does not understand itself in any way.