Thank you for your input, Matthew, John

I think this kind of concludes this discussion, since both of you are in favor of #2, and I was leaning to it as well.

As for FootnoteRef and Link, I'll be sure to test for regressions to the best of my ability. Not sure if I'll be able to fix underlying issue, but I'll have to try to know for sure.

One point I'm still not sure about is captions. On one hand, ImageCaption and TableCaption have semantic value, since they store more information than just Caption style. On the other hand, there *is* a style for captions already.

I'm thinking about having Caption in reference.docx and generating ImageCaption and TableCaption in the same manner as, e.g. SourceCode, basing them on Caption. This will have an added benefit of simplifying getting captions in docx reader (something I was planning to tackle in near future), since all captions, including Word-created ones, will be based on the same style. Any objections?