Random is peculiar. You can flip a coin many times and it may come up as heads for a long period before it turns into tails. The more you flip, the bigger the sample size and closer to the 50/50 you'd expect. All I'm saying is, if they had a bigger sampling size....
It could do that, but the vast majority of times it will converge to within +/- a few percentages points of 50/50 after a relatively small number of tries. That is basically the same principle in polling. You can have a poll that is off, but most of the time, given an appropriately sized random sample, the actual value will lie within the error margin.

