In your expertise, What separates the asociality that comes with mild/high functioning autism from that which comes from a coping mechanism for an unhappy childhood of someone who would otherwise be "normal"?
Whether that matters is a different discussion. One area it might is treatment (for those who want it). If two similar disorders have wholly different etiologies, there is a good chance that a treatment that works well for one won't work at all for the other.
Which is a prelude to my second set of questions:
How do you determine which is which in a diagnostic setting, where one normally has an n of 1 as an information provider?
How do you determine the need or kind of treatment and whether such treatment is truly effective or little more than a band-aid for what would be complicated sociological issues?
In what way does real or perceived agency (or lack thereof) have a part to play in asociality in relation to in-born or apparent autism itself and by what measure?
If your question is, how can we currently do this with Autism, the answer is, we cant. Yet. That is the whole point of performing basic research to understand the underlying etiology. If we stop doing this research, we'll never find out.
Wrt. your second question, a lot of medications are simply "bandaids", and that is indeed a problem. Adderall doesn't cure ADHD, Zoloft doesn't cure depression, Levodopa doesn't cure Parkinson's. They are acceptable for now because they offer some relief from the symptoms, but do nothing to fix underlying problem.
So for a concrete example Levodopa is given to both Parkinson's and Huntington's disease patients. However if we get to the point where we can reverse Huntington's disease using CRISPR to edit the trinucleotide repeat issue, should we expect this same method to work with Parkinsons patients just because the two disorders have similar symptoms? Absolutely not. Parkinsons has nothing to do with trinucleotide repeats. Which is good to know, so we don't attempt a highly invasive treatment on people who we know for a fact won't benefit from it (and may likely do more harm than good on them).