Yes. You can do this. I would make your event duration non-zero (e.g. 0.0000001) OR make it the duration of the actual events. This will allow you to combine the event and sustained regressors later. If you have a zero second duration event, then that alters the regressor to be non-equivalent to the sustained regressor due to a scaling factor.
The HPF for this design should be 2 time the block-to-block onsets of the same block if the blocks are not random. The default setting will be bad for your design.