Dynare output with loglinear option of stoch_simul


I have a very simple question but even after reading the Dynare manual, this is still not clear to me :

what is the output of a Dynare simulation performed with the stoch_simul command and the loglinear option ?

In my view, this was the same output as when log-linearizing the model by hand, that is all variables are expressed in log-deviations from the steady state.
However, when I run my model in levels with the loglinear option in the stoch-simul command, the mean values that I get for my variables are not zero, but are the steady state values of the logged variables.
Therefore, I guess that everything is expressed in log terms in the Dynare output and not in log-deviations from the steady state. So if I want the IRFs to represent the log-deviations from steady state after a shock, I should modify the output that I get with the loglinear option ?

I find this a bit strange as I thought that we usually log-linearize models to directly get log-deviations from the steady state and not logged levels.

Thanks a lot for any clarification !

With the loglinear option, Dynare conducts a Jacobian transformation to get the decision rules in logs instead of levels. It does not demean the logged variable, so variables will be in log-levels, not log-deviations from the steady state. However, for IRFs and moments this does not matter as they are always reported as deviations from the mean (at higher order, this makes a difference but there the loglinear option does not work).