To call autism a social construct is accurate according to the rhetoric / philosophy of science. "Autism" is observational, phenomenological, as long as it is defined in the DSM by traits and not a set of etiologies.
That does not mean "autism" is not real. It means we define it imprecisely and experientially by committees and standards organizations. We have no "one" definition even within research or clinical practices because some insist on the broadest definitions (the "spectrum" of which I am increasingly leery) and others want a return to Leo Kanner's rigid definition that excluded "full quadrant" IQs over 100.
If I assign the name "red" to a 650 nm wave, it is measurable and quantifiable. We can argue over the name chosen, but the wave itself is what it is.
Autism is not like "red" because we don't have measurements that are precise. We have DSM-III, IV, and V definitions that conflict. We have Wang debating Cohen debating Attwood over what should or shouldn't be Asperger's Syndrome (or is it "Disorder").
It's a label. It's nothing more than a checklist created by people to define what they themselves admit cannot be located and proved even within their profession. Clinicians argue with researchers constantly over "autism" because researchers are very rigid with the definition; many don't even accept the DSM-IV-TR vs. the DSM-IV because that alters the subject population for researchers.
The only label I reject that has been assigned to me is "mentally retarded." That was the ignorance of the 1960s/70s. But, the other labels represented the best of their time and the evolving DSM. If the DSM weren't socially constructed, my label wouldn't have changed every X years.