The shifting of fictional persons into the NAF is called into question by the firm stand of the FRBR-LRM report that fictional entities should not be represented as agents. In earlier posts I pondered whether a relationship like "Atributed to" might be defined as a creator/contributor relationship with a wider range than the one FRBR-LRM proposes for creator/contributor relationships involving agents. The use case in mind is the many Sherlock Holmes books in our collection attributed on the title page to "Dr. John H. Watson."
There may be another way to resolve this conflict between how we tag and how we use entity names. The "Attributed to" relationship might be defined as a type of subject relationship, in that it conveys a fact about the resource which is not truly a fact about its creation. Subject is an ever widening umbrella which (in MARC) has begun to acquire specific relationship terms and codes. Maybe it could also provide shelter for "Attributed to".
In that case, we might tag fictional person, corporate body, and place entities in the NAF using 100, 110, and 151 rather than 150 in LCSH, and in MARC authorities code these headings as 008/14=b, "not appropriate for use as a main or added entry." We'd be refining the scope of NAF and SAF away from use and toward entity types, which could be an easier division to grasp for users of the files and the AAPs.