rgatemod=2 and 3 have two resistances, a constant "electrode" resistance (for the contact) and a variable resistance.  rgatemod=2 computes the effective series resistance and adds a single extra node to the circuit matrix.  rgatemod=3 add two extra nodes, keeps the two resistors separate, and also connects the overlap capacitances (CGSO, CGDO, and I think CGBO) to the point in the middle of the two resistors.  I believe the argument is this is more accurate, but perhaps not worth the extra matrix row (which can make for a large matrix in big circuits).