If you're performing a true "dynamic seismic" analysis, then the computations involve a number of SRSS operations. Therefore, your counterpart is correct, you don't need to evaluate both the "+" and "-" directions. Also note that in a "dynamic" analysis any non-linear boundary conditions are linearized (that is why you set the 1st option on the Control Parameters dialog - "Static Load Case for non-linear Restraints").

If you're performing "static seismic", then you're simulating the dynamic event with a uniform "g" load. In this situation, the "+" and "-" directions could give you different results. However, this is an approximation, and really only valid if the response of the system is dominated by a single degree of freedom. (Of course you won't know that unless you perform both analysis and compare the results.)