Is there any chance the material you are compacting is breaking down to smaller sizes to a denser mix in the mold? Seems like a pretty high number if it is "boney". If the Proctor is on a changed material not representing the job itself, blame the lab test.
Try to find a container suitable for all of the material, since stone corrections sometimes are way off, especially if more than 40% stone. Then, compact the stuff in that container with the same energy per cubic foot that is applied in the Proctor mold. That would be your Proctor as long as you don't break down gradation to a different gradation even there.
Are you using a nuclear density testing device? If so, is is calibrated for the material you are compacting? Your problem may be that form of field density test. You may have to devise a different form of field density test, such as a small test pit lined with thin plastic and filled with a material of given density, such as water. If the water to the material when compacting doesn't help the density, your testing is likely wrong.
I loved to come on these jobs of a competitor. After the testing was found wrong, I got a new client.