Hi Claude,
The new GFS is now a non-hydrostatic model and its physics suite, like WRF, determines the amount of water there is suspended in the atmosphere in various forms. The models differ in how they determine this and may not agree on the amounts of water, but the representation as simulated radar reflectivity will most likely be using the same physical coefficients for the reflection of radar energy by the type and density of water in the atmosphere.
My guess is that WRF filters out anything below 0 dBZ as not relevant while the new GFS appears to cut off at -20 dBZ. I think that it is safe to assume that this is arbitrary and in any case, as Dominique mentioned, values of 40 dBZ and above are the ones that need to raise concern.
David