Rasmussen didn't do very well in 2012, but they are a reputable firm and they modified their survey technique. They don't poll over the weekends now. The astute observers here appreciated that there was a repeating cyclicality in their polling over the weekends that showed an improvement for the Democrat candidate. (My explanation was that on weekends normal people tend to be outside and away from home, not at home sitting by the telephone waiting for someone to call.)
There is more to "Live Phone", "Internet" and "Robodial" than the chart shows. IBD, for example, polls 65% cell phones. That gets expensive. And the analysis doesn't even get into who is doing the polling. College coeds at near-minimum wage who received an hour of training? Consider me skeptical that they are ferreting out accurate responses.
Gallup's original 1936 poll that put them on the map had a sample size of 50,000. The cost of doing something like that today would be enormous.
I can understand that news organizations tout their own polls, since they are paying for them, even badly-done polls are expensive and they confirm the bias of 93% of their staff.