That is a good point, I think PHD only goes up to 15 second exposure times. I would suggest 5-15 seconds is fine, unless you are trying to guide on a very bright star.
I expect some people guide with shorter exposures, but I doubt there will be anything to be gained from guide exposures shorter than 5 seconds on your setup.
The longer exposures times of course have the benefit that they allow you to guide on fainter stars.
A good way to find out is to experiment. You can take spectra of the same star with guide times varying from 1 second to 15 seconds and see if you spot a decrease in the signal or an increase in noise.