float CG_ModelScaling( centity_t *cent )
if ( !cent->currentState.iModelScale )
return (cent->currentState.iModelScale / 100.0f);
then look for cg_thirdpersonrange.value, and multiple the float by CG_ModelScaling( cent ) same with vertoffset
so i think in CG_CalcIdealThirdPersonViewLocation you would go to the bottom part where its doing vectorma, you will also first have to figure out the cent possibly centity_t *cent = &cg_entities[cg.snap->ps.clientNum]; but dont count on that working because as you see from the other parts in there, there isnt always a cg.snap.
VectorMA(cameraIdealTarget, -(thirdPersonRange * CG_ModelScaling( cent )), camerafwd, cameraIdealLoc);
pretty much do the same thing then for VertOffset ( CG_CalcIdealThirdPersonViewTarget )
havent really messed with it, their 3p code is pretty messy.