The two kinds of PDF tables
Tables in PDFs come in two flavors. Text-based tables store characters and their positions — the converter can read these but has to guess which characters belong to which cell based on spacing. Image-based tables are literally pictures of tables, usually from scanned documents. The converter runs OCR on them, and OCR on tables is much harder than OCR on paragraphs because cell boundaries confuse the layout analysis. If your table converts cleanly from one PDF but not another, check whether the bad one is a scan.
Merged cells and spanning headers
The most common conversion failure is merged cells. A header that spans three columns in the PDF becomes three separate header cells in Markdown, each repeating the same text. This happens because the PDF stores the merged cell as a single wide text element, and the converter splits it evenly across the columns it detects. There is no universal fix, but if you control the source, avoid merged cells in tables you plan to convert. Use repeated labels instead.