I'd say your experience suggests you can't tell imaged text from actual text

I've looked at a few of the free Google books, and a lot of them look like (poor) photocopies (i.e., with shadows and dirt) - and indeed they were merely scanned by Google and aren't "real" digital copies - a dead give-away that it's image and not text. In any case, elsewhere in the forums you'll find the common suggestion of running the PDF through an OCR process which will create an invisible layer of text over the image that will then allow you to highlight & select.

EDIT: When you export an eBook from the Reader, it creates images of the text. This is done to prevent full quality copies of DRM'ed text. Your PDFs, if not scanned from paper originals, may have been generated in a similar way.