This is a complex issue that depends on the exact role of the pavement in your model. If you're modeling the RUNOFF, you'll obviously need to use a reduced curve number. The specific value depends as much on your local stormwater regulations as the behavior of the pavement, since many localities assume degradation over time, and may be concerned about complete blockage due to ice and other factors.
In other situations you may be concerned primarily with modeling the INFILTRATION through the pavement. In this case you would use a high CN value to capture more (most) of the rainfall, and then route it to underground storage and infiltration, underneath the pavement. In this scenario you may have little or no above-grade "runoff" in the traditional sense.
Recently, I had a project utilizing infiltration pavers. This is the approach I followed:
1. acquired infiltration testing.
2. used a CN of 100 in all paver areas.
3. created a "pond" for the paver base with % voids to create storage.
4. used the resulting rates from the infiltration testing (a 1/2 safety factor design rate) for exfiltration outlet.
5. created a weir outlet depicting overflow from excess runoff (what could not be exfilrated).
