It looks like the new test tested 7MP, 10MP, and 16.7MP crops. That would be 135 PPI, 161 PPI, and 209 PPI. Assuming no other weaknesses in the imaging chain, that is, a high quality lens was used, proper postprocessing, and a sufficiently detailed subject (a landscape might have been a better choice than a portrait), etc., it would be a bit surprising if there were no identifieable difference in the 135 PPI image.
But he doesn't say that. He says only that only 3 people correctly identified all three. I'd really like to know the percentage of those that attempted an identification who correctly identified the 7MP image.
With regard to the larger MP images, I'm really not that surprised. 161 PPI is getting pretty close to the point at which it's supposed to be unlikely that most people would notice a difference.
Remember that you need to just about double the MP for the resolution difference to be really noticable. The difference between 10MP and 16.7 MP really isn't that great. But the difference between the 16.7 MP image (209 PPI) and the 7 MP image (135 PPI) really should be expected to be noticeable.