For a standard MaxDiff, it's a straightforward process to remove an item and re-proportion the probability-scaled utilities (from HB) at the respondent level. Then we have relative importance scores that sum to 100 among the remaining items.
My question is about doing the same sort of thing, only with anchored MaxDiff utilities, obtained via direct binary anchoring. Can (should) it be done? And if so, how? Should the anchored analysis be adjusted in some way before doing anything with the resulting respondent-level utilities?