I've tried out the test build in a few different scenarios and it seems to be working great, the hyphens behaves as expected!
The newly added unicode tag is quite enjoyable. I wasn't convinced at first, but it turns out to be explicit, yet clean enough (as in easily distinguished within a sentence), to be of use. We'll probably end up extending STM with a few custom tags/strings for our most commonly used 'special' characters, as per your suggestion.
Hyphen compatibility was our main issue right now, and that's clearly been resolved. Thanks a bunch!