I emailed them yesterday to ask for an update and they replied:
We can confirm that the problem is a real one - we overestimate the station error in some situations - but the effects are minor. We are working on fixing the problem and writing up this and the other corrections to the data that are described here:
So that's official confirmation then. Neat.
I wonder if 'writing up' means that there's a paper going to be submitted on this.