Its also worth noting, in addition to the point gojomo makes that party identification is self-reported and volatile (and the further, related, point that people both enter and exit the sampled population over time), so that the populations of "Democrats" and "Republicans" in the 2013 poll aren't the same as the population in the 2006 poll, that the questions in the two Pew polls that are the basis of that "metaanalysis" aren't about the same thing.

The 2006 poll question asked about warrantless wiretapping by the NSA (it specified both "listening in on telephone calls" and "without a warrant".)

The 2013 poll questioned asked about asked tracking call information under "secret court orders", which is neither listening in on calls or done without a warrant. They are two different things; it is a mistake to treat them as the same thing and consider different reactions to them as inconsistent.

(Though, given the specific differences, shifting from supporting the former to opposing the latter is something I find odd, but I suppose it could be consistent and not based on pure partisanship if it was grounded, for example, in a strong ideological opposition to the idea of judges overseeing executive national security actions.)