I see. To clarify, it parsed numbers in the pdf correct, but assigned them the wrong meaning. I was wondering if codex is better at interpreting non text data