One of the reasons we have lots of eBooks out there with errors is because a PDF source was used as the master document to make the eBook versions. I know of no program that can take a novel length PDF file and convert it without introducing errors. That's just how things are.
For example, I have read the first two Sookie Stackhouse books and by looking at the CSS, I se that they were converted from PDF. And because of this, there are errors. In most of the italics, there are missing or extra spaces at the begging and/or ends of the italics. Also, after the simulated small caps, the space was missing. Those sorts of errors are unacceptable. If I had not gotten the ePubs from the library, I would have been asking for my money back.
So please do not start with a PDF source ever. Take the PDF and delete it. Make It go away. If there is no existing PDF, there cannot be a convert ion that introduces errors that never should have been introduced in the first place. The only way to make sure a conversion from PDF is correct is to compare the PDF to the converted output and that means checking every word, every letter, every punctuation mark, and even every space.