Obviously, an entry in the LNAB field comprised of researchers notes which never was used by the person in question -- such as Bompasse-Bumpas needs to be corrected.
However, in some of the discussion I'm hearing a conflict between two motivations -- (1) what I understand to be WikiTree policy, that we try to give a person the LNAB that the person himself would have used at birth, and (2) the desire to give the person a name which resembles that of other family members, so that today's researchers will have an easier job of finding the person.
It seems to me that while (2) is a desirable objective, (1) is the policy that we need to be true to in cases where they might conflict. Say, Smith or Smythe, or Lewis and Lewes. I wouldn't change one to the other just to aid indexing, if the person himself/herself used the alternate spelling.