Parsing pdfs is very complicated when it comes to tables. There are cells sometimes that span multiple columns and while we as humans have no trouble understanding it, LLMs do not and parsing data properly is very important if we want good results. The documentation of Llama parse even says output in markdown doesn't do so well so try the output in html
But despite toggling on "Output table as HTML in the markdown", the resulting markdown I download does not contain any html. I even tried with multiple models, the result is always the same (Pipes and hyphens). What gives?
