Second Order Moments for the model in Aguiar-Gopinath

The latest versions of Dynare have as an example the model by Aguiar and Gopinath (2007). Well, the model has a clear mistake, like not including trend shocks (gt) into production function, but even after correcting for this, it cannot reproduce the second order moments in the original paper.
Here is the original file: … gtrend.mod

As suggested by Villemot, the stochatic simulations must be done with the order of approximation set at one. Still, I could not get the desired results. Any ideas?
I’ve also tried introducing observable variable which take into account that the variables are detrended by a nonstationary technology component, like:

yobs = y-y(-1)+g

We will fix the mod-file soon. Note that the mod-file does not intend to replicate the JPE-paper. For a true replication file, see

Can you contribute to the page in the ReplicationWiki about this study? … _(JPE_2007