I've seen that Agilent Technologies announces a breakthrough in high-frequency SPICE simulation technology for high-speed digital board design. It seems that they have just realized that simulation in the frequency domain is faster than that in time domain. I don't grasp the novelty of the news, because they have been doing so for a long time. However, the point I'd like to make is that a good model for RF design not only has to be accurate, etc.... but it also has to take into account the possibility of being used in a simulator oriented to frequency.
It seems then that in a good model to be used in an RF environment, we should be careful enough as to explicitly state the equations for the S parameters, and not only for the capacitances. Thus, we should present both versions of the model: one with the capacitances, and another one with the S-parameters. It would be then a question of choosing the most appropriate set of equations to simulate.