Read somewhere that his work was flawed because he was trying to capture and record the frames before they reached the rendering pipeline instead of after. There are many things that happen during rendering and a lot of those include throwing away frames. So as always take it with a grain of salt
Are you talking about the multi-GPU setups, specifically the Nvidia ones? The "flaw", as you mention, is that Nvidia apparently has a slightly different implementation of things in multi-GPU setups compared to AMD. It's a flaw in a sense that not all variables are held constant during comparison, but it's not such a big flaw that everything else posted about the article is automatically void.
And there are other things that are holding it back so to speak (like the use of IPS panels instead of TN panels, relying on software instead of hardware - like speed cameras, which he mentions), yet that does not mean that it already should have been taken with "a grain of salt."
All scientific articles published in academic journals all have certain "reservations." In your opinion then, those articles (all of them) must be taken with a grain of salt because they have "flaws."