That looks like a workable stopgap solution. But what happens if the document contains both footnotes and endnotes?
Actually, I think it would be better if this were handled professionally within Scrivener. But if that isn’t possible (because such ‘minor details’ don’t really matter), I still prefer to take the detour via Pandoc. That way, you also have the advantage of an app that does exactly what you expect it to.
If you say what formatting features you need we can suggest a Compile format for you (head over to the markdown forum for that). Do note that all publishing tools have caveats, Pandoc included. While Pandoc is extensible via filters, its native representation of a document structure (the abstract syntax tree) does not contain all elements all writers need. There is no perfect publishing system…
No one has said this, and I made a wishlist post to alert Scrivener devs to this potential small improvement…
I’ve discovered that RTF export works exactly as one would expect. As I don’t need to rely on DOCX for this, the problem is solved as far as I’m concerned. In my view, however, it isn’t for the company. Because if it works with RTF but not with DOCX (and only half-heartedly with ODT), it’s clear that they are well aware of how it should work, but that it has been implemented inadequately (one might even call it sloppily).
I am satisfied, however, to have found the solution myself, so I don’t wish to spend any more time on it for the time being.