I religiously avoid using the words "always" and "never" on the Internet. That said:
PDFs never convert well; they're always glitchy.
PDF was designed to maintain page layout, regardless of the display, whereas e-books are designed to have a fluid page layout per display preferences. The two goals are mutually exclusive, and therein lies the problem. There's no easy way to fix it, but there are some ways to minimize the effect. You can convert to HTML (or RTF) first and then edit the source. This is time consuming, but produces great results. If you use Calibre and you're willing to familiarize yourself with Xpath
http://calibre-ebook.com/user_manual/xpath.html and regular expressions, you can automate many of the corrections. Otherwise, you're just stuck with it the way it is. Sorry.