Would it be generally safe to assume that any command returned by exports(Typesetting) which "shares a name" with a MathML tag will do pretty much what the latter would do - or would it be safer to regard it as "suck it and see"?
That is true in most cases but not in all. For the purposes of displaying text in a plot, it should be safe enough to just experiment with MathML tags and see what gives you the desired result. The differences are more important if you want to parse the expressions you create (but that is not something that I expect you will generally need to do).
Do you happen to know if the Maple help for the Typesetting() package is going to be enhanced/extended to cover some/all of these undocumented commands, or are they destined to remain effectively "hidden"
There is no immediate plan to document these exports, as they are meant to be used internally. Most of the time, you can get the mathematical display you want without having to build such structures yourself. Of course, there are cases like the one brought up in this thread where it's not so easy to accomplish with the documented tools.
If it turns out that there is a need for better control of typeset expressions, then we would consider documenting these tags. It is more likely, however, that we'd try to provide tools to users to generate the typeset output they want. This situation is somewhat analogous to that for plot structures. Over the past few years, we've tried to improve the tools for generating customized plot structures so that Maple users don't have to work with the internal representation.
ETA: as mentioned, the Typesetting tags follow the MathML tags closely, though they are not identical. In acer's example, mn actually refers to numeric entities, so it would be better to use mtext for the titles, even though the displayed output is the same in both cases.