Hi Alena,
Are you getting the 3dFWHMx values from afni_proc.py? If not, what are you using to compute them?
In general, yes, use a mask when running either 3dFWHMx or 3dClustSim.
For 3dttest++, it may be preferable not to use a mask (or to simply run it twice, once masked and once not). It is helpful to have non-masked results, just to see if there are any strange clusters outside the brain. For example, many tasks lead to motion-based results at the edges of the brain, which blur into the brain and can look like valid results if initially masked. It is important to see and try to resolve that. If one doesn't know it happened, well, ouch.
It is simple to use 3dcalc to create a masked 3dttest++ result after the fact, too.
- rick