How to select the set of neuronal nets for S2 MSI data: C2RCC-nets or C2X-nets?

I have processed S2 MSI data for chlorophyll-a and TSM retrieval with C2RCC processor on SNAP.The processor provides 2 options, one is C2RCC-nets,the other is C2X-nets, for my comprehension the former is for relatively not so turbid and optically simple waters, the latter is for turbid and optically complex inland and coastal waters. But I haven’t got guidelines for how to determine the appropriate set of of neuronal nets . I tried both sets of neuronal nets for the same MSI data for Lake Taihu in eastern China, and the retrievals seems to have pros and cons, for chl-a tne C2X-nets performed better than C2RCC-nets,the spatial distribution and phenological characteristics are more realistic according to field truth. Yet TSM concentration of C2X retrieval may have valus more than 150,200 and even 300 g/m3, C2RCC retrievals are about half to 60% of the magnitude of C2X retrieval and the range seems are more ‘realistic’ compared to reports of other researchers.
Can any one give me guidance for proper selection of the set of C2RCC processor neuronal nets?Thanks.
the true-color composite of MSI data:
rgb-composite
the following chla and TSM are based on C2X set:
c2x-chlac2x-tsm
The following chla and TSM are based on C2RCC set:
c2rcc-chla c2rcc-tsm

For C2X, the maximum value ranges for the different IOPs for different sensors are these:

The document where you can find this information is the ESA Case2Extreme ATBD. I am not sure it is public yet. I will check if it is possible to publish it.

I guess you will have to check which one works better for your area.

Thanks to @ abruescas! I have seen the maximum values of IOPs for S3 OLCI is much lower than all others, does this imply no C2X neural nets for OLCI data retrieval? In my opinion this set for C2X is greatly expected since OLCI data are widely used for aquatic ecosystem monitoring of inland and coastal waters.

Not for this C2RCC version. We are testing new ones, but I cannot tell when we will have final results. In any case, I will have into account the demand for a C2X OLCI version.