Part of the value of a real pair programming session is brainstorming and discovering together. That part can't happen in a known problem for the interviewer. How does that part play in on the interview? I like how you use a reproducible problem for comparison, a lot of advocates ignore that.