Thanks @jmjones, great points!
Using speedtest sites to determine router throughput is not reliable because of several factors. We use professional-grade equipment and applications in a lab setting with no other applications running, no QoS settings, no firewall rules, etc.
If you really want to test the true capacity you should use iperf or nuttcp between two PC's or servers, one connected to the WAN and the other connected to the LAN. No internet connection is needed for this testing.
In the case of the Balance One we were originally able to hit 900Mbps + but we later discovered a weird oddity within the Apple/MAC iOS network stack that limited this. Hence we revised the throughput rating.
The point is is that there are several things that can affect router throughput but we do have to give it a rating.