i sort of agree with you, except you and me, we can not look out of the window without applying our "theories" to the world, so the notion of "i don't believe ..." can only go so deep. we are supposed to hold theories in an open space and improve them all the time. most theories will have to be translated (integrated) at some time, rather then fully abandoned. meaning that a statistical tool may prove it to be valid, but also proves it to be not a typology. this sort of differentiation is not based on actual "proving" of something, its the product of enriching one's set or perspectives and or sorting them, according to their own nature.
for instance we believe that chairs and desks are types of a single family (furniture). this makes sense so far, but if you would apply the most basic pattens of human thinking (eg the nature of holons) consequently to chairs and desks (and everything else), you would come to an entirely different result. concrete furniture is not a naturalistic typology, and it is not a holon, nor is a desk -if viewed on it's gross level of material nauture- a holon. but this perspective is only possible, once we zoom out so far, that we not only see furniture but the whole world, because only then we see what the relevant common denominators are, and what are random elusive lookalikes. (furnitures are artifacts. wood is close to a holon. a tree is a true holon. a forest is not a holon, but all trees on earth are a holon. the idea/design of a chair is a sort of a holon, but its of purely psychological nature, and not related in any natural/characteristic/structural way to the physical artifact)