I believe that your choice number 1, differences in statistical methodology, probably accounts for most of the difference in results. The raw data in all polls must be adjusted for demographics to correct for differences in proportion between the sample and the population at large. To cite an extreme and unlikely hypothetical example, if the poll respondents happened by a fluke to be 70 percent male and 30 percent female, the mean value of opinion taken by averaging all the responses would be misleading if males and females responded very differently. In that circumstance, the data for the males would need to be proportionately reduced for accuracy.

Obviously, it's easier to know how to adjust for sex differences than for more subtle effects of income, occupation, marital status, political affiliation, geography, and the like. Each organization does it differently, with Rasmussen tending to correct more in the conservative Republican direction than the other pollsters.

I don't know the exact figures but I have a sense that Rasmussen and other pollsters tend to agree more closely on clearcut choices such as a response to "who will you vote for?" than they do for varying shades of opinion that are hard to capture even if the polling tries to frame questions in a neutral manner, which does not always occur.