The Washington Post-Pew survey is detailed (through a link) in the article, specifically this question:
"Q: As you may know, it has been reported that the National Security Agency has been getting secret court orders to track telephone call records of MILLIONS of Americans in an effort to investigate terrorism. Would you consider this access to telephone call records an acceptable or unacceptable way for the federal government to investigate terrorism?"
58% responded 'acceptable' and 41% responded 'unacceptable'.
"Q: As you may know, it has been reported that the National Security Agency has been getting secret court orders to collect telephone call records of nearly all American cell phones in an effort to investigate terrorism. Would you consider this collection of telephone call records an acceptable or unacceptable way for the federal government to investigate terrorism?"
What isn't clear to me is if the phrasing has changed from the survey a month ago to the most recent survey. If the phrasing (of the entire survey because cross-question context is important) is not identical then the results just aren't comparable no matter what biases may or may not be built into the questions.
I've been working with some research companies lately and one thing which surprised me was the fact that surveys are often constructed with the sequence of questions carefully set up to test various messages. Unless you can see the survey itself, I would question the results.
For example, the question prior to 'How do you feel about the NSA’s accessing of telephone call records of millions of Americans through secret court orders' might have been 'Is it important that the NSA’s accessing of telephone call records of millions of Americans has stopped terrorist attacks?' This type of question can mightily swing the outcome, so if you don't have access to the survey you need to ask yourself if the organization putting out the survey has an agenda...