It would seem to me that masking works to a point (towards the "highly functioning" side of the spectrum), and once you get to the ways in which autism blocks ability to perform certain common behaviours and activities (e.g. non-verbal), the gender percentages should get closer again, because there's no masking to hide them or social expectations that they match. Is this the case?
It should be noted that this research sample, as opposed to many autism samples, includes autism with non-verbal, and severe intellectual disability phenotypes of autism, phenotypes not always captured in neuroimaging research of adults due to compliance issues (We collect neuroimaging data while participants sleep in the scanner).
