The example you mentioned shows a distribution for one test and then shows that an above level test yields a similar distribution among students at the tail. So it's not really a distribution with another hump at the tail, although I remember reading that someone else proposed a curve like that to explain the prevalence of high IQ individuals. In fact, I was thinking about this just the other day.
What if there were a gene that some small portion of the population had, and the expression of this gene were not on/off but governed by some complex interaction with other factors. Then the individuals who had the gene would express that gene along some continuum.
Here's an example: pretend the gene is for "extra long legs" (ELL). If we just take a random sample of the population we will include some individuals with ELL. We then measure everyone's legs and the results look like a bell curve with the ELL people at the far right end. If we then just take the ELL people and measure their legs, those measurements will fall on a bell curve, too. Their legs are all extra long, but they are not all the same length.
Intelligence is probably governed by a complex combination of genes whose expression is influenced by other genes and by environmental factors. So I wouldn't be surprised if the tails of the tails of the tails have humps. But the farther out you go, the smaller your humps will be relative to all the noise factors. So you really can't see what's going on. The test just can't focus on details that fine.
