Multiplying of expected variables

This is Professor Pfeifer answer for the exact similar question.
“If you use order=1, this is exactly what you should get. In a linearized version, certainty equivalence holds and the covariances drop out. The trick with the auxiliary variable is usually the correct way for higher order approximations. If it does not work, please post the mod-file.”
Many thanks
I got it.
let me rerun my code