The problem (as usual) actually lies in the fact that the PDF wasn't created with a consistent structure. It looked ok on the page, and whoever wrote it was satisfied with that.
I opened the PDF you gave in Acrobat X and exported it as XML (I'm sure there are other PDF editors that'll do the same).
I then opened the XML file in Notepad++, selected the actual text (between <?xpacket end='r'?> and </TaggedPDF-doc> and copied it. I then opened Sigil with a blank new document, changed to Code view and pasted this text into the body (i.e. select <p> </p> and hit paste), then saved the ePub.
The result is attached as example.epub. It's a lot better than the examples you show, but the poor structure of the source shows up in the lack of proper hierarchy and some paragraphs that are misplaced. It didn't need much editing to produce a corrected version, attached as example-Edited.epub. It'll require a reasonable amount of work to go through all the notes and check that the lists are ordered correctly, but I did this in about a minute, so it won't be too hard.
[Edit: If you
aren't using Sigil to create ePubs, now's the time to start. You do need some rudimentary knowledge of xhtml and css (no css was used here, it's all pure default styles), but it gives you far better control than anything else]