I agree that team cohesion, manager skill, and domain experience are significantly more important factors to the success of a project than physical proximity. A bunch of random people dumped into the same room will not outperform an experienced remote team. But that is not sufficient evidence that proximity is not a contributing factor.

To be more explicit: all else being equal (i.e. similar cohesion, skill, experience), teams that work together in a dedicated room seem to outperform expectations compared to similar teams that are spread across an open office or working remotely.

(Personally I did not notice anything close to the "doubling of productivity" claimed in the paper above, but it was sufficient to be noticeable even with a small sample size.)