Thank you Ric, the upper crash structure should help the rear wing (see my previous post) but is a temporary solution.
The definitive solution would be similar to JJR rear wheel arches (JJR car is my favourite).
Variante developed some details and your car (Brook) looks good too.
Mantium and TF could be good outsiders.
Talno has a couple of very interesting solution (maybe the most innovative car), but at first sight it would need some more wind tunnel testing.
UPDATE: I've just received the results. Quite happy, but I can't understand the difference of 900N between my local solver (OCCFD) and the official one. It seems that the rear wing works better on my PC