1. You show that you do not understand sampling. Since a single poll samples a "population," it cannot be precisely correct in determining the population average. Only a 100% poll of the entire population will be precisely correct. The advantage of sampling (hopefully random) is that the average of the entire population can be estimated, with some margin of error.<quoted text>
That is not the nature of statistical sampling, if a poll has an error of +/-2% that is the error. IF polls are differing one is wrong.
2. Depending upon the size of the sample, relative to the entire population, a "margin of error" can be calculated. Typically, a so-called "two-sigma" spread is calculated which means that there is a 95% probability that the true population average falls within the "two-sigma" error as added to the average value determined by that single poll.
3. You say that if polls are "differing," one is wrong. That is not true. Essentially, ALL polls differ because of the relatively small number of samples taken.
4. What IS true is that, if one takes SEVERAL polls, including polls taken by different organizations, and averages those poll results, that average number should statistically be closer to the true population average than the results of a single poll. That is the reason I like to look at the AVERAGE number of the polls which RealClearPolitics tracks.
5. I am amused at your article regarding polling. If you had read the link you provided, you would have seen that the historic polling was taken of MILITARY VETERANS. Be advised that military veterans do not represent the overall US population. The "military veteran" population is primarily male, primarily over 40 years of age, and is primarily composed of persons with "conservative" political views. Thus, the demographic of the article which you cite is not particularly representative in any sense of the typical US voter. We can reasonably expect that the "average" military veteran will vote for Romney. It will not make any difference.