On the Estimation of Treatment Effects with Endogenous Misreporting