Luis,
What is the assumed source of the mismatch loss? Variation in irradiance or module capabilities among the strings? Or something in the inverter itself?
There's a number of prior studies about mismatch loss from variation in condition of series-connected modules in a single string. I don't recall offhand studies of mismatch among parallel strings. The multi-input inverter model in pvlib assumes that all MPPTs in the inverter are identical, so if you are investigating variation among MPPTs, I'm quite interested in what you find.
Cliff