Text formatting issue due to which ai giving inaccurate responses

Maybe try the new version 4 with vision? Seems like you’d avoid all the weird formatting issues if you just upload an image of the pdf.