I've experienced some problems while trying to estimate the shading impact using 3D models on rooftops that have a "verge" (I'm not sure about the translation for this term, but I've explained with images, the "verge" is indicated by the red lines in the last image)
The system is on Porto Alegre, south of Brazil, around LAT -30.0277, LON: -51.2287
I'm not sure, but it seems like the shading calculation stops to work correctly if you place the solar panels below the height of the "verge". There are some weird triangular patterns. It's hard to explain, so I've taken some screencaps to demonstrate how the shading changes drastically if you change the distance between the solar panels and the rooftop surface. Also, I've taken some screencaps of the actual shades drawn by the 3D model, and it seems like the shading calculation is overestimated.