This is covered in the PDF
faq... This is a common occurrence with OCR pdfs that use images. It's true that if there is good OCR text that this is normally what's extracted, but there are many ways to define a pdf, and clearly whatever way yours is defined is not compatible with Calibre.
If the underlying OCR is good and you want that text, then the only option I can think of is for you is to use Acrobat Professional to try and change the way the pdf is formatted. There are a bunch of pdf 'optimization' options in Acrobat - for different pdf version compatibility/better compression - sometimes optimizing/re-compressing the pdf using Acrobat will make it compatible with Calibre's libraries (but not always).
Other
freeware/open source pdf tools may do the same thing, but I don't have any experience with those.