I think that there are real differences among gifted kids, but the further we get from the mean, the less useful tests are in determining who is what. There are so many ways to be gifted that it really starts being more about who matches the test and who had a good day that day. It's not really a linear construct (think about how we would think about an "athletic quotient" test, and how it might compare athletes who excel in very different sports).

Plus, the tests' reliability drops off (error bars go way up) outside of the "average" range -- they're just not designed for fine-grained measurements, even if it looks like they do. Hard cutoff scores are really hard to justify psychometrically, especially if standard error of measurement isn't taken into account.

And fundamentally, I also don't think the distinction between supposed levels of giftedness has much practical use. Yes, there are real differences among kids. But the precise scores or behavioral checklists aren't really how I recognize them and they aren't the information I need. When I'm doing an evaluation and making recommendations, I don't ever find myself saying, "This kid has an IQ of 130 so we should do X, but this other kid has an IQ of 140 so we should do Y." I take a much more multifaceted look at kids, their learning styles, their personalities, their resources, their environments, etc.