Acceptance estimation in log-linearized model


I have a log-linearized model with 90 equations (20 unit root equations), 16 shocks, and 47 parameters. All posterior shapes are normal except 2 stderrs Which have two tops in their posterior shapes. Interval, m2 and m3 curves converged for all parameters and stderrs except 2. Can I ignore the above problems and accept the estimation results?

Thanks in advance.

How can I solve the problems?
Increasing the number of blocks, change the acceptance ratios, change the mh_drop or …

You should investigate the trace plots why their distribution is bimodal. Is it the chains drifting or a true bimodality?