Currently, the consistency rule checks words. I would rather have it check regexps. There are so many compounds for ‘literatuur’ as well as ‘litteratuur’, which mean the same and are both correct, that checking all words makes the list very long, while checking using a regexp does not.
The same applies for mais and maïs.
I guess the routine collects all words in the text (or part of it), checks if any of the regexp pairs are in the words list; and checks if both words are in the dictionary (to avoid false alarms). Then it reports the inconsistency.