this charts (as many published polls) are based on very small samples. asking 1000-3000 people. yes those people are selected to sample the whole spectrum, but

this charts (as many published polls) are based on very small samples. asking 1000-3000 people. yes those people are selected to sample the whole spectrum, but if they are always the same people, you can not claim that they are representative anymore. as simply by asking the poll questions you are turning group of people who "don't care about the election" into the group "was forced to think about election and now have an opinion".