i don't know, i had a photo as reference (well, sort of), the caustics casted by the chair are exactly as the real ones (more or less). they are emitter size dependant and in this case i used a very small one.
the problem is we can't control caustics per object as yet, in this case the floor is not supposed to produce all those strong caustics (in fact it wouldn't in real life).
thanks guys. actually i started it as a sss plastic chair test, but at the moment it takes too long to clear up, more than 4 hours for a noise free image at this size.