This is the third paper in a series which combines N-body simulations and semi-analytic modelling to provide a fully spatially resolved simulation of the galaxy formation and clustering processes. Here we extract mock redshift surveys from our simulations: a cold dark matter model with either Omega_0=1 (tauCDM) or Omega_0=0.3 and Lambda=0.7 (LambdaCDM). We compare the mock catalogues with the northern region (CfA2N) of the Center for Astrophysics (CfA) Redshift Surveys. We study the properties of galaxy groups and clusters identified using standard observational techniques, and also the relation of these groups to real virialized systems. Most features of CfA2N groups are reproduced quite well by both models with no obvious dependence on Omega_0. Redshift-space correlations and pairwise velocities are also similar in the two cosmologies. The luminosity functions predicted by our galaxy formation models depend sensitively on the treatment of star formation and feedback. For the particular choices of Paper I they agree poorly with the CfA survey. To isolate the effect of this discrepancy on our mock redshift surveys, we modify galaxy luminosities in our simulations to reproduce the CfA luminosity function exactly. This adjustment improves agreement with the observed abundance of groups, which depends primarily on the galaxy luminosity density, but other statistics, connected more closely with the underlying mass distribution, remain unaffected. Regardless of the luminosity function adopted, modest differences with observation remain. These can be attributed to the presence of the `Great Wall' in the CfA2N. It is unclear whether the greater coherence of the real structure is a result of cosmic variance, given the relatively small region studied, or reflects a physical deficiency of the models.