I have written a program to overcome the problems with the Flanagan-Thorne procedure for transfer functions having very strong resonances by using time series of seismic noise.

Still, the results are not as sufficient as without these resonances but better than the procedure from Flanagan and Thorne. I uploaded a graph.