There was a third group where all members acted independently with no communication. The difference in likelihood of finding the optimal solution between that group and the one with intermittent communication was not found to be statistically significant in the paper, which is probably why they don't talk about it much. I think that knowing about it really does color the paper's interpretation, though.

It makes the cynic in me want to say that the real take-away here is that greedy search is more likely to get caught in a local optimum in a problem where the search space is not convex.

It's also worth pointing out that, to achieve P<0.05 on their main result, they had to run a LASSO regression with cross-validation. IMO, that counts as P-hacking.