I had a look at the file, but I have to admit, I didn’t find a solution to the difference between the produced IRFs and the IRFs from the paper.
The BankControl.mod was written by Kevin Moran after the publication of the article. In a way, it is a replication file of the original version that was not written in Dynare. For the version with Bank Capital Channel he comes relatively close, but still the dynamics are less persistent.
For the case of no Bank capital he recommends in line 176-180 of the file:
% In some experiments, we want to leave mu_ss unchanged at 0.025 (to compute s.s.) but have mu = 0.0 for the linearization (to produce “WITH BANK CAPITAL CHANNEL” responses and “NO BANK CAPITAl CHANNEL” responses). In order to do that, need to disable lines 81-84 in the “steady_.m” program of your dynare build
Now with Dynare 4.5.6, the steady_.m script only has 55 lines, so it is not clear to be what he was aiming at. Either way, he introduces an inconsistency in the calculation of the dynamics and the steady state!
One way I see to to replicate his approach would be to keep mu=0.025 in the calculation of the steady state, linearize the equations by hand and replace mu in the linear system by 0. This would probably do the trick.
If you find a better way to get closer to the result, let us know. In the meantime, I will add a .txt-file to the replication package that highlights the difference between the results in the paper and in the MMB. Thanks for pointing this out!!