I have a feeling that most of the peer-reviewed research on protein-intake and muscle gain is seriously flawed, largely due to some of the reasons best explained here (free article)....
http://medicine.plosjournals.org/perlserv?request=get-document&doi=10.1371%2Fjournal.pmed.0020124The biggest problem I have seen with most studies on this type of subject is very low sample sizes, even before dividing them into treatment & control groups. I don't believe that protein intake is nearly as important as what common bodybuilding sources claim it to be (2+ g/lb bodyweight), but I would bet the effect is somewhere lower than that; in other words a small to medium sized effect from more protein. Most studies simply do not have the statistical power to find this, even if the data actually show it. I don't think I've ever seen any article on this subject that uses alternate methods for inference, i.e. bayesian or even bootstrapping or simulations. The mags typically stress 2+g/lb, but they make most of their money from advertising and usually protein powder is a huge chunk of that. These magazines completely rely on this source to stay afloat, and if protein powder is found to be overhyped, what are they going to replace it with in advertising? Even w/ creatines, NO2, thermogenics, etc., the protein powders clearly rule the field for ads.
Realistically, I agree with this article's author and think that many of the disciplines that use classical inferential stats to draw conclusions are putting out at least 50% false findings. In some of the fields that use human subjects (psych, exercise-related sciences, drug-testing), I wouldn't be surprised if about 75% of what we think we know is false.