Jerry from Ectaco forum said me that the PDF document contained image masks that were not downsampled.
It seems that the "image masks" are a way Acrobat has of producing holes inside of transparent objects for other objects to shine through.
They tried to reoptimize without success.
Anyway he will pass this problem to the technical staff.
As workaround he suggests to convert it in jpeg but I think it is not a good think. I instead prefer to do:
pdf2ps problem.pdf p.ps
ps2pdf p.ps without-problem.pdf
New "without-problem.pdf" will be bigger than "problem.pdf" but it will be a pdf with text and image and not simply a image-only pdf (so you can perform a text-search, for example)