Selecting the correct HRF


I am performing 1st level analysis of task fMRI in a clinical population using FSL Feat. One option for each EV is the convolution to apply; that is, which HRF to select (e.g., Gamma, double-Gamma). I understand that the HRF can vary on a patient and brain-area basis. I also understand that adding the temporal derivative to each EV can help if there is not strong prior beliefs about the HRF to use. However, I would like to understand the pros/cons to each function.

I have read some papers relevant to my task and they say “canonical HRF”, which I presume means the gamma HRF. I also have read several tutorials on using FSL Feat (for example, this one), and they generally advise on selecting one function, without much justification.

Are there best practices on how to select the proper convolution to apply?