It's not based on registrations. It's based on Rasmussen's party identification poll, which is conducted every month.
However, even Rasmussen says that it's not valid to use this poll to adjust others that are sampled separately, or with different methodologies.
I think that most of the polls are skewed. It's not necessarily intentional, but promoting them as accurate without pointing out the obvious oversampling of Democrats is intentional. A reputable pollster would never do it.
But, it's hard to get an unskewed poll these days. Many Republicans are disgusted with the media, and refuse to participate in the polls. Others are simply too busy -- they aren't at home to answer the phone. The only people that participate in polls these days are the ones with nothing else to do.
And even then, you have no way to know if the respondent is being honest. I think you can imagine a number of situations where you would be subject to retribution if you said you wouldn't vote for a particular candidate. Put yourself in that situation: if you got a phone call from someone claiming to be a pollster, how would you respond?