Kari'smom's question, "have you ever stolen anything?" was a control question which are as broad as possible. They are intended to concern the examinee so much (which is what the examiner wants), that the physiological reaction indicates deception to the same degree as an untruthful answer to the relevant questions will. That's why control questions are so broad. They are questions that will worry the examinee either because they really cannot answer "no" or because they are worried about answering "no". The irrelevant questions, "Is you name Susie?" and "Is today Tuesday?" are expected to not have a physiological response at all and are used as the baseline. This is all explained in the links I provided a few pages back. No Deception is Indicated when the physiological response to the control questions is much greater than the physiological response to the relevant questions. I hope that makes sense. Deception IS indicated when the physiological response to the control questions is like the physiological response to the relevant questions.
Something about this bothers me and I'm glad that polygraphs aren't usually admissible in court, because I think it's quite likely that many innocent people are more worried about responding to the relevant questions, regardless of if they know what happened. Everyone with half a brain can tell that "do you know what happened to Deorr" is The Question that you shouldn't fail. And if you care about Deorr at all, the question might make you more anxious simply because you are once again reminded that Deorr is missing and something dreadful has happened to him.
I think they can be useful in getting the people to respond to interesting questions and eliciting confessions in the hands of skillful investigators. But the thought that innocent people would be stressed about responding to questions about, say, some irrelevant crimes that no one is investigating, and remain carefree, physiologically unemotional and unstressed while responding to the relevant questions about a crime that they're overall very emotionally distressed about - it just doesn't seem obvious to me. If you read about the science behind the polygraph you might be nervous also because you know there's the possibility of getting a false result and getting suspected for a crime you didn't commit.
Sure, if people who changed their stories multiple times fail the polygraph, I'm often inclined to think they could have failed for a reason. But the whole theory behind the polygraph examinations seems like it's on a shaky ground and the reliability evidence is not too impressive so there would be a number of false convictions if it was used in the court a lot.
ETA: I do believe the machines are pretty reliable in detecting heightened physiological responses that are commonly associated with nervousness, it's just that the physiological reactions and nervousness are not always consequences of lying.
Some form of guilty knowledge test might work better but I'm not sure it could be used here because the examiners need to have some information that only the police and the guilty party know (eg. what the murder weapon was, or some details about the crime scene), and if they don't know what happened to Deorr, it might be impossible to design such a test.