Some very important points to remember when considering the veracity of these polls.
1. This and any other survey assumes the person being called is really of any party affiliation which is a subset of considering which ones are chosen for the publicly published sample.
2. Not to accuse Rasmussen, but consider every vote is in a database. I can setup a routine from a database for a predictable result by actually sampling 2500 people, assuming their party affiliation is true, and running the routine to find the correct number of each party on a percentage basis who answered the question a certain way with a selected total of 500 as what this sample is numbered.
3. First and foremost, Polling COMPANIES (key word here) are in the business of selling polling to those who are looking to either find the truth or as in many cases, create a result to make news in an effort effect public opinion, not to find it. This is why there are such differences with internal polling which we never hear about since the owner of the results are the ones who payed for it. Likewise, those that are made public are most often commissioned to create news in an effort to sway public opinion.
Not sure I got your #2: are you polling all 2500 or randomly chosen 500? What routine are you talking about? Some selection algorithm? Based on what?
Either way, heavy manipulation of survey population or resulting data would be suspicious and detrimental to the poller’s reputation, so they wouldn’t risk it. I am afraid, they are close to truth.