You always have to look at the whole data set from multiple angles. Lets take
Do you support or oppose - requiring background checks for all gun buyers? 91% said yes, 8% said no
and compose that with
Do you believe that if there are background checks for all gun purchases the government will or will not use that information in the future to confiscate legally-owned guns? 48% said confiscate, 38% said will not
Lets put those two things together.
Even if we assumed that all the people against background checks believed they would lead to confiscation, and peel them off the list, that still leaves 40% of the people polled who are both in favor of checks and also believe it will lead to confiscation.
If you think that checks will lead (not might lead, will lead) to confiscation, why would you be for checks… because you want the guns confiscated.
If they think the guns should be confiscated, why are we even debating them about background checks. They don’t care about background checks. Those people are not trying to keep guns out of criminals hands, they are trying to keep guns out of everyone's hands. So this 91% stat is misleading at best, and useless at worst.
Either that, or there is some serious cognitive dissonance going on and if that’s the case, the poll is invalid on its face.
In the spirit of “trust but verify” here is the original poll that everyone is citing.