I agree with IFRs. I think this is also something that should be made more clear in the standard. And it's not clear to me if this was actually the intent of the writers, or just the result of poor wording or oversight.
See the definition of Dead Load in 5.2.1. Also note the usage of "nominal" (= "as-built") roof thickness in App. F, sections F.1.2, F.4.1, F.4.2, etc.
Note that when calculating uplift forces on anchor bolts, the corroded thicknesses are used.