Strange posterior densities

I am trying to estimate a search and matching model of unemployment, a very basic version as in Lubik (attached). I use observations on output, unemployment (rates), wages and vacancies. I detrend the data prior of estimation and add a constant term in order to get all positive numbers. However, the results are very bad…
Could someone help me in order to understand what am I doing wrong? Also considering that with the simulated series I obtain good estimates, what can I do to improve the fit?

code_estimation.txt (1.64 KB)
Lubik.pdf (211 KB)

There’s a new company profile out there. Looks interesting: … namic.aspx