Chosing the Berendsen coupling constant in MD simulation

I am trying to simulate a charged particle system where the system is equilibrated to a desired temperature using Berendsen thermostat. I have seen that for different values coupling constant of Berendsen thermostat, the static structure of the system ( using Radial Distribution Function ) comes out differently. How do I ensure which result is correct ?

If your system is properly equilibrated, the simulation settings a proper so that there is good energy conservation, and the is g(r) properly converged, there should be no significant differences in the g(r) for different thermostat relaxation constants when chosen in a meaningful range.

For a given system, how can I choose the range of relaxation constant meaningfully ? What are the factors that govern the value of relaxation constant ?

Please study the relevant publications that describe that thermostat algorithm.

But as I mentioned, there are multiple other factors that are more likely to be the cause for inconsistencies.