My expectation is that you will get full speed on the B One when you hardwire it to your Netgear and perform a hardwired speed test from your PC. If true, then it will make sense to investigate what is happening between the ONT and the B One when the Netgear isn’t in the middle. At that point, I would probably place a switch between the ONT and the B One as this has fixed some issues between Ethernet chip sets in the past (certain Peplink products have this issue with Starlink’s USB Ethernet adapter), although the switch trick hopefully doesn’t apply to the B One.
Getting a report back from you on the CPU utilization when the problem is occurring may shed light on the issue. How does the CPU usage look when the problem is occurring connected directly to the ONT versus connected to the Netgear?