I would be very gratefull if somebody could clarify a bit the forecasting part from reference manual, i.e.:

Is the model first simulated and then the last value of simulation period is taken as the forecast origin? Or I got it wrong?

How would one proceed to compare forecasting preformance from DSGE with some empirical model (e.g. VAR), i.e. how should forecast origin be specified?
Any help (also in terms of some references to lecture notes, slides etc.) is more then appreciated.