Amazing! This question must not have occurred to the authors of the code or the editors of the De Garmo. The part-measure repeat sign clearly defines the meaning of the pseudo-eighth notes of the group, so I see no reason not to group before the repeat even if the next note is a true eighth. In any particular instance, try it out and see if you can detect any ambiguity. If you can misunderstand, don't group in that case. I tried a number of hypothetical circumstances and couldn't find any such ambiguities.
I'll add this to the things to consider in updating the code.