There seem to be a few issues surrounding OOP that produce biased "evidence" for OOP.
The first is collection-orientation. Data-centric languages and tools fell out of favor after OO became "in style" in the early 1990's. These data-centric features filled in many of the gaps that a procedural-only language would have. Thus, by killing the data-oriented languages, OO can brag that it has features or capabilities that "procedural languages" don't (or are at least harder to deal with). An example would be using a ControlTable instead of case/switch statements. (C's ridiculous "break" switch-statement is also language-specific anti-evidence.)
The second is "clean code". Newbie programmers were often told that OOP was the best way to create "clean code". Thus, those interested in producing cleaner code gravitated toward OOP. However, this increased the ratio of bad programming done in non-OOP language. It is not that OOP is inherently "better" at code organization, its just that the hype drove those interested in clean code, such as having less repetition, into OOP.
Third is procedural techniques atrophying. Features such as those found in ImprovingProceduralLanguages will not get considered because vendors are too busy adding OO-centric features.