I wouldn't necessarily disagree, except that "Correct Answer" does not always mean a technically perfect, complete solution to a problem; it's often a move to close a discussion by noting the best applicable answer.
But if you insist on a "correcter" answer, how about this—
- You insist you need standards-perfect EPUB2. You haven't said why, even though some discussion of your project and needs might lead to a roundabout answer that can't be reached by just insisting.
- You haven't said why you can use only EPUB2, a rather sketchy standard some seventeen years old and outdated by a standard that's now some thirteen years old. Again, knowing the why and what could lead to workaround solutions; just saying "because" is not productive.
If we assume you absolutely have to have fully validated EPUB2 output, you have the choice of—
- Using ID, exporting to EPUB2 and accepting some elements that don't fully validate. This is often no crisis, especially if the elements are flagged on an "info" or other non-failing basis.
- Using ID, exporting to EPUB2 and reverting to EPUB2-era methods of editing the exported file to make it fully compliant.
- Using a a prior version of ID if there is one (you are insistent that this is a v19.5 problem) to generate your EPUB2 files.
- Using some other workflow entirely to create your EPUB2 files, as was done for many years before ID came along.
But, any notion that Adobe is going to rush to "fix" unknown "problems" in the export module for this ~20 year old and almost wholly obsolete standard is misplaced.
(ETA: Given the vast variation in EPUB tools and production, and the near total lack of precise adherence to the standard across the entire spectrum, expecting either perfect output from ID or that Adobe would post a notice or remove the feature is... peculiar.)
Until and when you're willing to share more about your project and needs, especially an explanation for your insistence on 'perfect' EPUB2, that's as correct an answer as you're likely to get.