For my simple use case (just counting pages), switching to smalot/pdfparser solved this issue completely. The same PDFs that were failing with FPDI's compression errors worked perfectly. Obviously different library with different capabilities, but worth trying if you're hitting these FPDI limitations.