Of course what you say is correct, but I find it impossible to look at a set of statistics (even basic stats such as percentages of votes) without dissecting and analyzing said data for representativeness, variance, relevance, reliability, and a few other esoteric things that might only be of interest to a hunk of germanium.
But your link is good too; reminds me of the classic stats issue regarding US Presidential election projections from 1932 or somewhere around there. A very large telephone survey was conducted by a newspaper, and the results were landslide conclusive. Problem was, only about 15-20% of households owned a telephone back then.
You can guess the rest.