You have literally described all surveys, good job.
It isn't like people who run surveys don't know this and take it into account, there has been plenty of math that has been sussed out to try to remove the kind of bias that can happen in self reporting.
There's reams of science dedicated to this problem, it isn't new. I'm tired of seeing this argument shat out without a shred of consideration for scientific and mathematic history of taking surveys.
That being said, there IS a more accurate way to define Apple usage among teens and that would be to survey the school districts.
"How many students do you have? How many devices connect to your network wirelessly? What percentage of those are iOS?"
That still wouldn't be 100% accurate due to teachers and staff bringing their own devices, but the students vastly out number staff so the variance could calculated.