1) The fact that polls can be constructed to show anything doesn't doom them to show nothing. Use the same wording, &c., across different times, regions, demographics and you can get interesting results.
2) The fact that polls test what people think is socially acceptable rather than what they "really" believe is most likely a feature, not a bug. The former probably has more explanatory value.
3) An intense qualitative investigation can get data with more depth and less breadth than polling. But can it get closer to people's "true" opinions that are otherwise mediated by the construction of the poll - context, framing, &c? Does that question even make sense? I doubt it.