This is the problem. Whatever you're polling for, this is a fallacy. I'm actually inclined to agree with your initial response, but your way of deriving it is not valid.
Polls can and often are accurate representations of the general populace without the requirement that 'everybody must be polled'. The poll in question may not have been for numerous reasons- and even if it were accurate I would consider its relevance low, but your criticism that 'it's not accurate because not everyone was asked' demonstrates the understanding of statistical significance of a third grader.
Don't compare me to a third grader just to prove a point. I've taken several lessons on polling in my AP Government and AP English classes as we often were given them as an option for supporting essays and research papers. However, we had to be very careful about using them because they aren't accurate nor precise. Sure they'll give you a general answer for the general public. However you must ask yourself, who is this general public? Polls, even when taking random participants, are going to have varied results depending on location taken. One location is going to have a different norm than the other. If you're trying to use this poll to validate a different location....
Here's an example of a simple poll: Do you like Apples? Yes or no.
I can answer it in any number of ways:
- Yes because I consume them a lot
- No because they're sticky
- Yes because they taste good
- No because they leave you feeling hungrier
- No because I'm alergic
- No because I might have just bitten into a waxed/rotten apple
- Yes because I just read a study on their benefits
- No because I just read a study on their acid
- Yes/No because I hate polls
- Yes because I want to fit in but I really don't like them
- No because liking apples is too mainstream
- No because screw polls, that's what
Times that by your sample size and you're left with a mess, not to mention data from people lacking in experience. Sure it might be generalized but it's not accurate. Polling is only accurate when you know you're getting truthful answers from small sample sizes (the small sizes such as a classroom) that'll only apply to that group. Class A and C might love apples while Class B hates them. Generalizing this data to say that the school must love apples is inaccurate and wrong.
If you have the option for a different source, use it, not the poll. I know you agree with it's low relevance but they shouldn't be used to represent stuff.