Hi Fiona,
I'm reading the guidelines and thinking that you might be swimming in the opposite direction. You should be using the FWE corrected p-values, and should avoid using unprincipled approaches (I'm using same words as in the paper).
For a permutation method it doesn't matter whether a cluster extent threshold is applied before or after the permutations have happened. All that such a threshold will do is to erase from sight the small clusters, without affecting the p-values for the other ones. This doesn't sound good right? Indeed, it shows how arbitrary the very existence of the cluster size threshold is, let alone the choice of a value. So, follow the guidelines and don't use it.
All the best,
Anderson