The secondary processor exception exists for legitimate reasons. If you have some machine with 100% open source boot firmware, Coreboot, etc., that's still worth certifying even if some Ethernet chip turns out to have firmware inside it.
What Purism is doing is moving memory init code away from the CPU on which it would normally and most naturally execute (in terms of engineering design decisions) for the express purpose of moving something they can't actually open out of the scope of RYF. And are then brazenly admitting how they are gaming this in the above blogpost.
Most people would expect "open source firmware" to mean "open source memory initialisation", because we understand that's something the firmware does. Moving this onto a secondary CPU is literally no better than proprietary firmware. It can't be audited, and if it's able to initialise the memory controller, it can be expected to have full access to the system. They're practically carving out an Intel ME-like boot processor here, with much of the same disadvantages and concerns.