I have only a few "branches" (alternations) for the forward-looking assertion. But how would I put a capture group there? In any case, I've made several attempts at this now with no success. For some reason the match always ends at the earliest possible alternation, apparently recognizing the (.*?) in the separate branches as being equal. Going to an unqualified greedy capture would not work either, as that would potentially slurp entire paragraphs instead of sentences until a match would be found--combining chunks that should remain separate.

For my part, any way that will do the job works for me. I don't mind, for a one-time-use script like this, writing pages of messy looking code that is inefficient, ugly, whatever--if so long as it would get the job done! So having a large number of capture variables is a non-issue. I've sometimes gone above 30 on those for a job of this nature! But in this particular case, that is not going to solve the problem.