While underlying rock is important in generating soils, recent glaciology is critical and often mixes things up substantially. Only in tropical soils and very thin soils are you likely to always have a strong relationship between underlying rock and soil.
I don't actually recall hearing about glaciers and their effects on soil forming, or how soils are chosen in game.
I actually meant that the soils should have their own stats for mineral and pH levels in the raws, as well as stones, where stone layer stats are only read and used in cases where you are growing plants directly on muddied rock.
So, then, loamy soil would tend to have some fairly good soil quality initially, while sandy soil might be more difficult, but any decent soil would probably be better than the initial conditions of growing something straight on top of stone, where you might need to raise some plants specifically to break minerals out of the stone, and split the stone.
As for soils, being as they are in the raws (and hence, totally interchangeable with any soil that has similar tags), I would expect the only differences between most soils that the computer can read are in whether soils have the [SOIL], [SOIL_OCEAN], [SOIL_SAND], and [AQUIFER] tokens, so all soils with the same tokens would be indistinguishable, and hence, just totally chosen at random when generating biomes. (So Peat, Silt, Loamy Sand, Sand Loam, and Sandy Clay Loam are all functionally identical in the computer's eyes, as are all forms of sand.)