Did you run your FEAT analysis with the optimal/custom HRF convolution option in the FEAT model setup? There is also an issue with filmbabe in 5.0.7 due to it's use of a deprecated executable, so a version of FSL prior to 5.0.7 should be used.
> If filmbabescript is no longer maintained/functional, I am happy to develop and share code for doing a hierarchical Bayesian constrained FLOBS fit with HMC methods. Of course the variational method used by filmbabe will be vastly more efficient, and I doubt anyone would want to wait for whole-brain voxel-wise HMC, so I'd prefer to use filmbabescript if it is intended to be working.
> Just wondering about the status of that tool if anyone knows...