There are many models of neurons, at many levels of detail. I think that the Neuron program uses the finest detail of any existing software.

I see the primary purpose of a simulating a nematode as measuring how well such models actually work. If they do work, it also lets us estimate the amount of detail needed, but the first question is whether these models are biologically realistic. An easier task would be to test whether the models accurately describe a bunch of neurons in a petri dish. The drawback of such an approach is that it is not clear what it would mean for a model to be adequate for that purpose, whereas in a organism we know what constitutes biologically meaningless noise. Also, realistic networks probably suppress certain kinds of noise.