While I'm just as lost as most when it comes to pfSense and HFSC, something I read seemed to indicate that, on a realtime service curve, d is the maximum time elapsed before it gets its m1 or m2 rate fulfilled. For example, if you had a game that required 500Kb bandwidth with a good ping of 30ms or lower, you would specify m1 = 500Kb, d = 30ms, m2 = 500Kb. I don't even know if you need to specify m1 in the rt case where burst is not a requirement.
Please bear in kind that I don't know what I'm talking about, and the above could be complete nonsense.