While Calibre is useful for some purposes, I'm not a fan of its conversion system for anything other converting from another reflowable format which has already been properly edited. If you look a bit closer, there are lots of errors in the ePub, including some serious paragraphing mistakes (e.g the paragraph starting "In the mean time" in the section marked "Why Dragon's Fury?"), which are not, in this case, the fault of the PDF. It's also introduced totally spurious characters which do not appear in the text at all. I found the original PDF, and its paragraphs are all correctly tagged, so something in your conversion process messed up badly. With a little bit of effort you can certainly get something that is considerably better.
The best starting point is to export the PDF as html from Acrobat. This preserves as much information as possible. The next step is to deal with all the inline styles that are present (i.e. <SPAN style="font-size:9.9pt; font-weight:normal; color:#000000" >), so open it in a capable text editor (you can do this directly in Sigil, or use Notepad++ first) and use the replace function to change those to a named class, preferably one with a name that makes sense (e.g <SPAN class="dedication">), then define that class in the css at the top of the file. You often find that needless tags have been entered, and it's a good idea to strip those out too.
At this point you can define your class styles and decide how you want the body text to look, as well as the headings and any additional text styles. If you're working in Sigil, the next step is to check that the section and chapter headings are tagged consistently (<h1>, <h2> etc) and mark them as places to split the output file (in book view, set the cursor at the start of a chapter heading and click the 'insert chapter break' button).
Once your table of contents is correctly defined you're pretty much done. It's always a good idea to scan through the preview to check for any loose bits of text that haven't been styled correctly, then you can export an ePub to see what it looks like.
For the few times I've had to do this sort of conversion, I tend to use Word once the inline styles have be processed, as it's a bit slicker, then I use Atlantis Word Processor to output the ePub, but you can get a perfectly good result just using Sigil.
|