I think the key issue here is the difference between the type of content and delivery option used in the studies. The original Long Tail hypothesis was tested on a large online digital audio collection, while the new research uses Netflix data on movies ordered via mail. In the first case, the risk to consumer to waste valuable leisure time is essentially non-existent. Audio download takes very little time and it's easy to try a song and dump it after a few seconds.
It's different with Netflix movies, though. The waiting time for a movie is at least one day, and if the user plans for a having a good time, a risky bet on an unproven movie may not be worth it. Since we know that people, in general, are risk averse, choice of a better known movie is justified from a behavioral point of view.
In any case, the Long Tail theory seems to work best when "total consumption costs", including access and trial time, are close to zero.
tags: control, scale, selection, risk, payload, distribution, entertainment