I just did a quick check for worldwide witch doctors (all 4 regions) for the current non-season using
http://ranks.zeroempathy.org/e12.wd
This only lists the top 200 era WD clears (Range GR 143-122). Of these, 30 were paragon 4.5K to 5.5K. Their average clear was 128.0. A similar result is obtain if one only considers 4.8K to 5.2K paragon (n=12, average = GR 127.6).
For necromancers using the same method, the average was 121.6. (GR range 136-120 for entire top 200)
For DHs using the same method, the average was 126.4 (GR range 141-125 for entire top 200)
Their data looked at the leaderboards in 4 regions. Therefore, one would anticipate that these leaderboards have additional clears that did not make the top 200 worldwide for ~5K paragon players (i.e. equal to or less than GR 122 for WD as an example). Therefore, the average for all WD players ~5K paragon who made their region-specific leaderboards must be equal to or less than 128.
Nev mentioned that they averaged regions first and then got a global average. As such, one could imagine that the numbers can be skewed if one region is an outlier.
These numbers in the blog post table seem puzzling even more so in light of: