Huh, somehow between hitting multiquote and responding I got your argument confused. Mea culpa. In my defense, I did just read 10+ pages of assorted arguments. In my detriment, I didn't re-read the one I happened to quote.
I recall now what I wanted to say. You postulate that you can't tell place a line because the difference between steps is too small to notice. That you can't say when 'genius' stops because the difference between steps is too small, so since you can't tell the difference between 20 and 18, you similarly can't tell the difference between 16 and 18, and so on down to between X-2 and X. While I fully agree each step lacks a clear definition, I disagree with your argument that there can be no line. Let me try to explain.
Let's agree that a +1 difference isn't noticeable. Let's further agree that there is some place where cumulative difference does become noticeable. For the sake of argument, let's say that's at a +4 difference. If two characters have a +4 difference in bonus, that will become noticeable at the table during normal play.
If that can hold, then I can place a line of difference. If I can say that 20 INT is the smartest of geniuses (barring oddities), and that it is the top end of the class I will call genius, then the bottom end is the point at which I can tell a distinct difference, ie, below 14. Once the difference becomes noticeable, then the classification can change, even if each step isn't distinguishable, the total can be. I can place a line with the total -- the point at which it noticeably becomes different.
Now, we can argue as to where that line may be. It may be more than a +4 or less, but the point is that you can draw a line. Much like I can draw a line between nighttime and daytime even though I can't distinguish the minute to minute difference in light during dawn or dusk(presuming I also can't see the sun).