Dear Jordi,
There is no specific motivation to keep the ratio you are referring to between density and temperature a constant. As mentioned in the post by Anton the densities in the different layers need to be accurately implemented to make sure the cross sections are correctly converted to mean free paths for proper particle transport. Then, the stopping power in the different layers is corrected for the density effect, first through evaluation of the Sternheimer-Peierls formula at Normal Temperature and Pressure (NTP) and then corrected back to the pressure (and by consequence temperature through the ideal gas law) you provided using the MAT-PROP cards. In this way it is certainly possible to simulate atmospheres with the temperature ranges you quote without errors, keeping in mind of course that in having the correct density profile is the most important.
If you are worried about possible boundary effects when going from one layer to another: of course this is inherent to the model in itself since the atmosphere is anyway sliced up in layers of with different properties. So in short I see no obvious reason why this would not be possible.
Hope this helps,
Andreas
(PS: I believe the ideal gas law equation should be T = P/(rho*Rspec)