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.