ID-exported reflowable epub rife with fragment identifier errors upon validation

EPUB validation is, by and large, a technical exercise in accomplishing nothing much. It's like blueprinting a car engine when you rebuild it: satisfying for the perfectionist and granting bragging rights, but it doesn't really make the car run any better. Validation is perhaps more useful when constructing EPUB files using the primitive tools and builders and editors wherein stoopid mistakes can be made; not so much with files exported or converted by high-end tools or proven utilities.

The problem is that EPUB is still a horribly implemented standard and the plethora of readers varies widely in interpreting it. Having that perfectly validated file doesn't mean any greater number of readers (software and human) will be happier with it.

If you're stopping with EPUB (e.g. not continuing on to Kindle or another final format), the single best "validator" is Thorium reader.. If it presents properly on Thorium, you're done. You don't need to keep fixing technical violations that probably won't make the book look better on less-rigorous readers.

The developers of Thorium are doing what all the participants in EPUB publishing should have done long ago: uphold the defined standard, not flood the field with loosely-implemented, "easy to use" tools and methods.

Discovered this same problem. I don't know about Calibre or Sigil, but with some kind of HTML editor you can rectify this. I used Atom, though I also had to download this app for zipping/unzipping epub files (I'm on a Mac)

In the book's xhtml files, you'll see InDesign creates unique IDs for the endnote hyperlinks in the text, like this:

<span id="endnote-001-backlink"><a href="#endnote-001">1</a></span></span>

This then corresponds to, obviously, the endnote itself:

<span id="endnote-001"><a href="#endnote-001-backlink">1 - Lorem ipsum etc</a>

So when you export one individual chapter, this all works perfectly, but when InDesign joins these chapters together into a Book, the IDs start clashing. The first endnote link in Chapter 2 gets given the id "endnote-001-backlink-1" and href="#endnote-001-1", as it should to keep the IDs unique. BUT the actual endnote itself has the id "endnote-001" again - it doesn't correspond to the links in the text. So the endnote link, and the backlink, can't find each other.

In a text editor, you can edit the endnotes so that those IDs actually match the links in the text. Depending how many notes you have, this could be a total headache, but it did the job. This is definitely a bug with exporting .indb files that Adobe needs to sort out. Or if there's an option within InDesign to skirt this, I'm not aware of it.