Flatness is a geometry that is in reference to itself, thus dependant on itself. If the item in question cannot support itself in a geometrically stable state, then the initial reference cannot be established. Enter restraint conditions. Is restraint feasible? No? What about Datum Targets? No?
If the part cannot be supported to establish the primary plane and then measure deviation from that plane, due to flexibility based on gravity then I submit:
1) You most likely cannot measure flatness on this part. Even the most sophisticated optical technology would fail in reliability because the surface form would be dependant on the form of the part support .
2) Your customer will also not be able to measure flatness on this part as it is defined in geometry.
Have you contacted your customer with this problem? Sometimes a customer will have a method they use, whether correct or incorrect in the ideal sense. If they do, you can match their method, you'll be on the same page and you live happily ever-after.
Short of all that, as much as I despise hearing these words;
If they cant check it, they cant reject it!.
Seriously, I would contact them for clarification. Maybe the part used to be aluminum or..., and they never took that control off the print. Just a thought.
Regards,
Steve