Well I tried a pi filter but this did nothing, I tried an old RS mains common mode filter, nothing, I have a common mode choke from Wurth just arrived I will try but I think we are just barking up the wrong tree with this power supply. Tellingly we get higher peaks with no load than with a light load, there is "excess energy" in the supply that has nowhere to go.
What common mode choke from Hurst? I'm talking about getting some big honking ferrite toroids or clamp-cores and running the wires through those, perhaps three turns, and perhaps several cores in series.
Again, I can't see what you've got and perhaps you've done everything you can, but a PI filter has capacitors to ground (or at least to a common point), and where you connect that common point can make a big difference. Just *try* the simple common-mode choke, an effective one, and see what happens. And back up when doing the radiated measurements. As I've mentioned, stuff you can pick up with a near-field probe may not be the same stuff that is actually radiating. It may be the same frequency, may be the same actual signal as what is radiating, but until you know *how* it's radiating you don't know much.
Aluminium reflects more than it absorbs so any RF energy is getting plenty of opportunity to get onto our wires after a filter.
"After a filter" has me worried. Aluminum is an excellent shield, and you're not going to fix this with a magic absorptive enclosure. If you put a filter at the supply, then run wires from that to a hole in the Al box you are going to be very disappointed.
Stop looking for the production solution. First find out what you have to do, then figure out how to do it practically.
I'll back off now and wish you good luck. I'm always happy to give advice.