The statistics in this paper are... not robust IMHO. They do so many comparisons between the groups and never mention any multiple comparisons correction.

Interesting idea, I am not convinced right now - but let's see if it reproduces.