Error with estimation: qz_criterium too large

I have one more issue with Alpanda et al. (2010) model. After I fixed the problem of misspecification I embarked upon estimation (actually mode finding) and I got the following error: c:\dynare\4.6\matlab\dynare_estimation_init.m: option qz_criterium is too large for estimating/smoothing a stationary model . If your model contains unit roots, use option diffuse_filter, suggesting that unit root is present in the model.
However, I calibrated model to prior mean and no unit root neither collinearity was found.

I attach dataset (xls as *.mat file I use is not allowed), code for estimation and calibaration.

Thanks in advance for advice.
Alpanda2010_model_c3.mod (5.83 KB)
dataset.xls (53.5 KB)
Alpanda2010_model_mf1.mod (7.58 KB)

You only provide a different datafile here. My guess is that the problem comes from you specifying

options_.qz_criterium=1;%Tolerance for unit eigenvalue 

in your code. For estimating a stationary model, options_.qz_criterium must be smaller than 1. It is not allowed to be equal to 1.

Thanks for your reply. I missed the fact that this line remained in the file. I removed it and added measurement error for foreign inflation to avoid stochastic singularity and it works now.