The Autism Phenome Project is predicated on the idea that there may be some underlying homogeneity so that we might identify clinically meaningful subgroups within the autism "spectrum"/"diversity", which may help us develop better targeted interventions, support services, etc.
Great question.
I suspect it will create confusion to use the same word for a multi-dimensional meaning, especially when there is preexisting literature about the "autism spectrum" that used the single-dimensional meaning. I think it would be ideal to use another term.
It seems that the thing we're talking about is "a cluster of imperfectly correlated traits". Googling a bit, I see "syndrome" as one possibility (although it may carry the connotation of a disease, although that may be inevitable with autism). "Trait cluster"?
(If one wants to dig into the light metaphor: The thing you're looking at is a beam of light. You use a prism, and that splits the beam into its constituent frequencies. When you shine the light through a prism onto paper, the resulting image is a spectrum. But the original light beam itself is not a spectrum. Thus, the most appropriate way to use "spectrum" in connection with autism would be for someone to take a test that measures several different traits, and then the test results are the "spectrum". I'm not going to try to fight for that definition, though.)
I haven't kept up with the current state of understanding, but now I am intrigued and I wonder how I score on those other dimensions they now asses.
Instead of one score, think of it like a cumulative score. You have, let's say, 7 factors all scored from 1 to 5. If the score for being diagnosed with Autism is 21, for example, there are many ways to achieve that. Three areas with a 5 and then a single 3 in any other area would have you diagnosed. Similarly if you score 3 across the board.
Basically, they're saying when you hit that minimum overall score, your factors put you that far divergent from most people that it's at least noticeable.