fortunately, this is yet another case where indirect specular isn't needed.
maybe somebody will do a render to see if one bounce of a directional lite off the umbrella
is alot slower to calculate, but gives the same result as, no lite bouncing of the umbrella,
which is set with reflector face either ambient or translucent = (white, 5 or 10) e.g.

it would also be instructive if somebody could do the same render in P8 and PP2010 to see
how much better or worse tone mapping makes it look, compared to simple GC.