Plotting Problem by using Dynare

Since the model in level just gives response of variable in level to one standard error error terms, I plan to draw some IRFs rescaled by the correponding steady state level. I used the variables generated by Dynare but it happens that those values is not the same as what is shown in figures generated by Dynare, some of them even have opposited signs.
You can check whether Inf_epsi1 is the same as that shown in figure: response to epsi1 and Interest Rate and Average hour.

That really drive me crazy. Can you explain why is the case.

I attached my mod file here.
TSJH_try.mod (8.5 KB)

This was indeed a bug in version 3 (not in version 4). I have corrected it and the correction will be available in the snapshot for version 3 from 02/05/07 on.

Thank you for reporting it.