I agree with most of everything you have said. That is common in many places, but exceptions are also regular enough to give good hope, and moreso in the post-graduate arena from what I have seen. Most of the people I know in this field work on or with theories that are more socio- than neuro-, but the balance is different at each institution.

Fortunately I see some very high quality and rigorous research techniques applied (with much less handwaving away of neuroscience) among my friends, but I don't need to look far for contrast. I don't work directly in this area, but am present for the pub discussions as it were.

However, I would say that statistical sophistication often seems to rely on the happenstance of cross-domain expertise. People only have so much time, knowledge, etc -- just the usual faults of humanity and organisations.