Impulse (singleton) dummy in GMM

[dvd.ceccarelli]

Dear guys, I have this problem. I have seen in the GMM residuals that there is an outlier. I have decided to insert an impulse dummy in the estimation. But when I have inserted it also as instrument, the Hansen J-test is implausible (p-value=1). There is a theoretical problem in using impulse dummy in GMM? Thank you.