Yes it is still difficult.
I have established a model (i.e identified 3 main modifications of the standard one-diode model) which works quite well, and matches the real modules behaviour with a very good accuracy.
The problem when using this model in PVsyst is to find the good parameters, as the information provided by the manufacturers is usually not sufficient.
However the default parameters proposed by PVsyst seem to give good results.
See my paper Performance Assessment of a simulation model for PV modulesof any available rechnology
, ( 25th EPVSEC – Valencia, 2010), available on our site http://www.pvsyst.com