" You are exactly correct. Ever get graded on a "bell curve"? If you were born between the years 1952 and now (was that Horace Mann?), you have been graded on a bell curve.
And yes statisticians can easily manipulate the data to make the poll be whatever outcome they want it to be. "
Gaussian distribution has absolutely nothing to do with cherrypicking data. Pollsters manipulate "data" by using different connotations when asking the questions. If you're reporting a simple statistic such as:
Yes? No? Undecided?
with a margin of error, the statistician cannot manipulate it without simply misquoting the statistic. The statistic may have been manipulated to begin with by asking a loaded question, but once the question is asked, you cannot really change the result if it's a simple question (though combining questions with interests groups to find "trends" can lead to manipulation). In any case, a bell curve has no relevance on anything other than the margin of error.