PDFs that are mostly images don’t parse correctly at this time. This is typical when users scan paper documents to PDF.
We are working on a fix.
Makes sense. I was trying to print to PDF to force a more unified output but I guess that survives that process.