> It seems a block HRF would be more appropriate conceptually but I can't find it actually
> used anywhere so I'm wondering if I'm missing something with this PPI step?
No, a block HRF would not be appropriate. An impulse response (a response from a presumably instantaneous stimulus) is used in the deconvolution step to find the neuronal events. In the case of a block experiment, it is expected to find consecutive events within each block based on the deconvolution of the kernel function (impulse response). If you use a block HRF, you would most likely end up with one event for each block, which is associated with the stimulus timing (onset time) of the block.
Gang