It's not just a limited sample, it's a biased sample. That makes it worse than useless for gauging public opinion.
You wouldn't keep contact with the kinds of people in Iran that respected and supported Khamenei; they wouldn't want to talk to you and you wouldn't want to talk to them. This is the equivalent of people who think Thing is popular because everyone in their social circle loves Thing, and thing Other Thing is unpopular because everyone they know hates Other Thing. Worse than useless.