Mapping offsets to matching tokens

Sorry, I must have confused myself. Offsets are utf-16 in both cases and they seem to be reported correctly regardless of analysis chain. I'm not sure how I got it in my head that the offsets were utf-8.