Thank you to all who responded to my recent question about Portuguese names. I've adjusted two existing NARs and created three new ones, based on your posts as well as correspondence with all five authors. All of them save one advised me in words to the effect of entering under the last element of their name.
The one who didn't, who happens to be the leader of the group, was one of the two already established. Curiously, her existing NAR (n 2004012947) uses the prefix as the entry element. I left the authorized access points in this and the other existing NAR unchanged, while adding information in other fields. The other NAR (no2004044937) has a simple surname, forename AAP format which the author endorsed as in use in his publications, though not in the one that provoked this discussion: OCLC 952384213 LCCN 2016956701 Software developers as users (still in process in LC’s catalog).
What would have helped is having the equivalent of AACR2 rule 22.5C4 in RDA. Without it, I found instead F.11.8, which leaves unsaid how to deal with a compound surname that follows the prefix. And without the benefit of cataloger's tradition, of which AACR2 has now become a part, a literal application of that instruction can lead to an incorrect formulation.
Perhaps someone reading this message will start a process to take care of this situation. I don't expect to see the AACR2 rule in RDA, but it isn't needed. Rather, adjusting F.11.8 might save someone else time wondering about this question. If I need to write directly to another address about this possible change to RDA please let me know.
Sincerely - Ian
Cataloging and Metadata Services Librarian
George Mason University