The previous theory was wrong because someone made it up on the basis of nothing. This is a very common event.
It is precisely analogous to the "dispute" over Pluto,¹ where the only argument was "I was taught in school that Pluto is a planet, which means it is true that Pluto is a planet". That is also the only argument for differential readability of serif vs non-serif fonts. It shouldn't surprise you that it turned out to be wrong.
¹ (The conclusions are not analogous - "Pluto is a planet" is not capable of being true or false when the definition of "planet" is up for debate. "Serifed fonts are more readable [under condition X]" is capable of being true or false. But the arguments are identical.)
There are more, but OP claims to dispute this study among others, and while I am very curios and have added it to my reading list, if somebody has already read and can highlight what was wrong with all the previous studies, that'd be great (that was the ask all along).
For instance, in this very study, Verdana (a sans serif font) is ranked best due to low mental effort, but an average reading speed is much better for serif fonts: I would like to see even larger x-height fonts than Georgia — eg. Baskerville — tested as well. Perhaps sans serif fonts are taken down by the worst ones.
I guess I'll need to read the OP's full 160 pages to get my answer.
It is precisely analogous to the "dispute" over Pluto,¹ where the only argument was "I was taught in school that Pluto is a planet, which means it is true that Pluto is a planet". That is also the only argument for differential readability of serif vs non-serif fonts. It shouldn't surprise you that it turned out to be wrong.
¹ (The conclusions are not analogous - "Pluto is a planet" is not capable of being true or false when the definition of "planet" is up for debate. "Serifed fonts are more readable [under condition X]" is capable of being true or false. But the arguments are identical.)