If you model your floors as membranes you will find that the distribution of reactions in the columns matches that if you performed the calculation manualy using tributary areas. The difference in reactions is due to the in plane stiffness of the floor when modelling using shell elements.
As I understand the intention of the stiffness modifiers is to account for reduced stifffness of cracked structural elements. I would have thought that the flexural stiffness of members (i.e. bending properties) only would need to be modified.
