Yours is a very good answer. Unfortunately, i think that the research on that field is almost at the beginning.
Anyway, many works in literature tried to characterize the Massive MIMO channel through the well-know "Winner" channel.
If you need further information, you could have a look on the papers of Emil Björnson and Luca Sanguinetti. They are two of the major expert (in my opinion) about MIMO and Massive MIMO.
There are not so many large-scale channel measurements for massive MIMO so it is hard to say what is the most reasonable model.
Most people that work with mathematical analysis in massive MIMO assume simple Rayleigh or Rician fading models, mainly for analytic simplicity and since most measurements show relatively similar behaviors.
When making more realistic models, the WINNER model is a good starting point. Some researchers have extended this model to handle physically large arrays where different antennas may see different scattering clusters. One example is:
S. Wu, C.-X. Wang, H. Aggoune, M. M. Alwakeel, and Y. He, “A non-stationary 3D wideband twin-cluster model for 5G massive MIMO channels”, IEEE J. Sel. Areas Commun., vol. 32, no. 6, pp. 1207-1218, June 2014.
In general Emils answer is correct. But as it is predicted that 5G Massive MIMO communications would be used for small cells, small as public place and halls, the plane wave channel models would be not so accurate to measure the performance of MIMO. It would be better if we could use a channel model which incorporates spherical wave modeling.