What's in a poll?

I was in the library the other day, browsing through some news mags. I looked at the latest (May 24) Time/CNN poll, because it had some bad news for the president. One of the questions (the last one, I think) went something like "Given the revelations of prisoner abuse at Abu Graib prison, are you more/less inclined to support the war?".
Interesting question. The answers are really interesting!
Sixty one per cent. said the news and the pictures had no effect on their opinion. Thirty one per cent. said it made them less supportive of the war. Eight per cent. said it made them MORE supportive of the war.
Here's a clear example of what's wrong with polls: I really want to know what it is about a picture of pvt. Englund holding a leash attached to a cowering naked Iraqi man that makes someone say - "Yeah! Now I see what this war's all about! We should be doing a lot more of this!" That's 8% folks, almost one in twelve, who thinks this sort of thing is a good idea.
Well, this has really made me think. And try to figure out what is going on. So far, my best idea is that this shows that about one in twelve people polled have exactly no idea what they are being asked. This might explain other seemingly anomalous poll results. Another possible answer is that about one in twelve poll respondees are sado-masochists. I hope that's wrong. Perhaps one in twelve regards this all, as Rush Limbaugh suggests, as nothing more than just a few little good-natured franternity pranks. Considering the body count, and the number of prisoners sent to the hospital, I have to think that Rush must have been in a really tough fraternity!
Time/CNN: How about some followup questions to that one-in-twelve? I really would like to know what kind of people I'm walking among.