Pew treats the opinions of the 5 million NRA members + the opinions of the 9 million self-identified transNRA members as = the opinions of NRA members.
Personally, I like to see the survey sampling method (how random), call opening spiel ("Hello, I am calling for Xyz, We are doing a survey on Abc, May I have a minute or two to get your opinion on Abc? Thank you. As you know, Abc is a matter of national concern ..." ), the actual questions, the actual answers. Not the political spin of the polling firm or the sponsor of the poll.
Wright, Rossi, and Daly "Under The Gun" (Aldine, 1983), Chapter 11 looked at two polls, one sponsored by the National Rifle Association and the other by the Center for the Study and Prevention of Handgun Violence and showed how results can be skewed.
Quote a review: "They discussed the data showing that gun owners - rather then being a violent, aberrant group of nuts - were at least as psychologically stable and morally sound as the rest of the population. Polls claiming to show that a large majority of the population favored "more gun control" were debunked as being the product of biased questions, and of the fact that most people have no idea how strict gun laws already are. As the scholars frankly admitted, they had started out their research as gun control advocates, and had been forced to change their minds by a careful review of the evidence." -- Dave Kopel, Independence Institute
I believe that a study that asked basic questions about existing federal gun law, then followed with questions about more gun control demonstrated that support for more gun control exists among those least knowledgeable about existing gun laws.