IRFs from linear model differs from first order approximation od the NL one

Dear All,

I have linearized a simple RBC model but when I cross-check against the non-linear version of the model I get different IRFs.
I have checked the steady state relationships for some variables and they coincide between the 2 mod files.
What can I do to solve this issue?
Many thanks in advance.

RBC.mod (1.3 KB)
RBC_linear.mod (1.7 KB)

My guess is that you still have a mistake in your linearization