Heres what I found on calibres Conversion section. It sucks because I was doing the same thing as you bud. Now I have to do some HTML. The PDF input to EPUB would make life a lot easier for me. That's ok im fine with the HTML.
http://manual.calibre-ebook.com/conversion.html
As the converison site says.
Some limitations of PDF input are:
Complex, multi-column, and image based documents are not supported.
Extraction of vector images and tables from within the document is also not supported.
Some PDFs use special glyphs to represent ll or ff or fi, etc. Conversion of these may or may not work depending on just how they are represented internally in the PDF.
Links and Tables of Contents are not supported
PDFs that use embedded non-unicode fonts to represent non-English characters will result in garbled output for those characters
Some PDFs are made up of photographs of the page with OCRed text behind them. In such cases calibre uses the OCRed text, which can be very different from what you see when you view the PDF file
PDFs that are used to display complex text, like right to left languages and math typesetting will not convert correctly
To re-iterate PDF is a really, really bad format to use as input. If you absolutely must use PDF, then be prepared for an output ranging anywhere from decent to unusable, depending on the input PDF.