In softly broken SUSY, the bare mass parameters may be specified at e.g. the GUT scale, and then we can run these down to another scale using RGEs, similar in form to the RGEs for gauge couplings, with a 1-loop and 2-loop differential beta function. Once these parameters have been run to the desired scale, tree-level physical masses are computed via diagonalization of the mass matrices depending on how the mixings were determined. Next, additional accuracy is gained by working out the loop corrections to the physical masses with self-energy graphs.
I am confused by why the physical masses that you determine from diagonalizing the soft parameters are considered tree-level masses if they are determined from 2-loop RGEs. Could you have determined the physical masses at the high scale and computed the self-energy loop corrections there, and then renormalized the physical mass down to the lower scale?