There's a problem with a purely arbitrary approach to the meaning of every word.
The most basic problem is that you understand what I mean. But a more advanced problem is that you know what I mean because you ascribe to each word non-arbitrary meaning and group these together as an intention.
In other words, on a certain "Wittgensteinian" level, there are rules to this language game. That's not a complete disproof of your point, but it is a necessary propaedeutic. Or to put it another way, it's arbitrary that the word knowledge is the world knowledge (or any way of saying it any language, e.g. Wissen, chishiki, le savoir ), but the basic concept (assuming a language has such a concept) is non-arbitrary.
And in this particular instance, the idea of the concept is that it's the concept of rightfully identifying the way the world is.
Thus, the word/concept unlike some other words contains an implication. The implication is that if you have knowledge, then you have knowledge of knowledge. But that if you lack knowledge, then you lack knowledge of knowledge.
For most other words and concepts, I could know it or not know it. I either know what a chocolate cake is or I don't. And I can check this by seeing whether I know it. But for the nature of knowledge, the thing to be checked and the thing that checks it are the same. Thus, the criterion problem.
We can now in a sense return to an objection you raise: what if "knowledge" is not what we think it is, i.e., what if this word is just like all the other words and concepts. In that case, I think your 100% correct, and that there's no special problem. But then the question is whether we've solved the criterion problem or just moved the goal posts.
And I think a major problem is going to be that it seems like to make sense of language and the way we think, you're going to have to accept there are at least a few words and concepts that aren't just arbitrary in what they do in the system. And one of them is going to be "did I get this right?" where this is the relation between the words I use and the world. That relation is going to suffer from the criterion problem since its condition is identical to its fulfillment.
Or so I would believe.
The school of thought that rejects that there's an underpinning to which our language attaches is called "post-structuralism." (though to be honest it's not that different from structuralism). You could look at Derrida, Kristeva, Barthes, or others if you're interested in denying the criterion problem and working with a wholly semantic concept of language.