[crankypants]

Roccokain, I just went through this yesterday. Here's what I learned.

1. All ids must be unique within an epub book. Don't bother making IDs unique within all <img> tags, or all <h1> tags, just make them unique within the whole book across all tags.
2. All ids must begin with a letter.
3. All filenames in content.opf or toc.ncx, like XHTML and CSS files, must begin with a letter.
4. The mimetype file is case sensitive, it MUST be in all lowercase, with no byte order marker, and no carriage return at the end of the single line inside it.
5. The mimetype file must not be compressed within the EPUB, and MUST be the first file. Only info-zip does this correctly AFAIK.
6. Just because an ereader can read your EPUB does not mean it will pass validation.

And last, the BookId value of the id attribute in content.opf is case sensitive. Here's a correct tag:

<dc:identifier id="BookId" opf:scheme="ISBN">1223456988</dc:identifier>