Choosing the TURF reach criterion leads to multiple possibilities indeed. The first choice threshold is one (and this is available in our TURF software). A continuous probability of choice method is also available, which considers in a weighted way (proportional to probabilities of choice) how all the items reach you. And, there are middling choices as you mention (top-2, top-3 reach), which our software doesn't do.
Personally, I wish we also supported the top-2, top-3 criterion, so our users would have yet another option to try for reach criterion. However, I feel very good about the proportional probability of reach approach, which does a great job leveraging the strength of the probability scores we can derive from a MaxDiff experiment (using the logit equation).