I had a test that I "cheated" on by using other clues. It was a facial recognition test. I was shown a bunch of cards with photos of people. I was then shown a book that had photos two by two. I could see what the majority of people had picked due to indentations in the cardstock from which the book was made. I recalled 10% (crazy moustache man, for example). I used my educated guess for the rest. After the test, I simply told the person what I did and how it would skew results. I was praised for using an adaptive technique.
Sounds like you basically did the same thing by reasoning out what the answer should be, not how you see it. You see vacant eyes and makeup and can reason that it is "desire" (or that's what she's trying to achieve). To you, it just looks a bit what? Empty? Since people in real life aren't just photos that we can study, that kind of deductive reasoning may get you a little farther along, but as a facial expression changes you may have to re-reason it out again. Doesn't mean you interpret eye expressions the same way as a NT or that your reasoning isn't serving you well. You've found a way to work around a problem. You're clever. That's OK - I'd just explain it to the tester.
Oh and by the way, one test usually doesn't make or break a diagnosis.
_________________
Apologies if I sound judgmental, preachy, dictatorial, offensive or overly rigid. Constructive criticism via PM is welcome.