I don't believe that's literally true, some all legal and correct Japanese unicode sequences will be transformed under at least some of the unicode normalization forms, no?