Please help! strange smoothed shock and histogrical shock decomposition

Dear Johannes,
Thank you for your previous help, I am grateful.
We have now estimated Christiano, Motto and Rostagno’s model using Chinese data, we can run the program, but the smoothed shock for interest rate and historical shock decomposition are very strange, we tried to modify but unsuccessful, could you please have a look and give us some guidance?
The first attachment is our mod file, second and third one are mat files containing initial parameter values, fourth one is data file, fifth one is the dynare output that I printed, sixth one is the picture of smoothed interest rate, the remaining ones are pictures of historical shock decomposition.
Thank you very much.
Best regards,
Jesse and colleaguesUS_CMR10faData26.rar (5.8 KB)policy_param.mat (1.9 KB) ss_CMR_FA_US.mat (4.7 KB)rawdataChinaDSGEnewnewnewnew.mat (8.3 KB)pdfOutput.pdf (83.8 KB)SmoothedShockPart1.pdf (77.1 KB) SmoothedShockPart2.pdf (19.8 KB)OutputHistoricalShockDecomposition.pdf (71.7 KB)

My guess is that it has to do with the observation equations and the means of the observed variables. The large jump in the smoothed shocks in the first period suggests that the model jumps to a higher level and then remains there.