Since the others tackled polygraph's uselessness, I want to comment on another angle:
I think fundamentally in such a case it will be easy for you to convince yourself that you're telling the truth in the moment you say it.
After all you are telling the truth to a version of the question, and you only have an assumption that the questioner means a different version of the question. Even if it's a good assumption, nothing in particular makes your version worse, in fact you could argue it's better.
That combined should make it easy to mentally gloss over the contradiction. So I think your physiological reaction will be indistinguishable from telling the truth on control questions.
Seems to me like they were already aware when writing this Press Release, they mentioned the policy on DSD: