21 days is to 95% probability of infection/non-infection. If you wanted 99%, you'd go out another few weeks. Conversely, if you wanted a more-or-less of 75%, you'd estimate 16 or 17 days.
I suppose probabilistic trade-offs govern a lot of CDC pronouncements--like the contagion likelihood from people who have a low-grade fever. But those guesses are even sketchier than the incubation ones, because they'd require a lot more data and viral measurement experiments than have taken place.
Is 75% good enough to allow exposure of your children?