To make this research more interesting I'd like to see:
1) repeated queries from the same user. Do the results stay constant over time or do they change?
2) comparisons to the same experiment run against e.g. Bing or DuckDuckGo.
It seems to me that some variation in results is to be expected because of users hitting different backends which might be at different stages of index rollouts. Similarly, response times of different backends matter. If for example the video results don't come back in time you'll end up not having them in the result set.
Lastly, the insinuation of the article is that "unbiased" search results are clearly preferable. I'm not convinced. I for one like that STD for me is associated with the C++ standard namespace (which I search for all the time) rather than sexually transmitted diseases (which I luckily don't have to care about as much).