1640's tests do not appear to account for where the center of mass is. That is a fundamental problem with getting raw values from that methodology for CoF (particularly when using an actual FRC bot as opposed to a block of steel, for example). If CoM is biased forward for Step 3, it will cause a higher CoF raw calculation, whereas a high CoM (z-axis, height of robot) will cause a lower CoF raw calculation.
However, it is good for a rank order when comparing wheel types, so long as the wheels are the same size or the chassis frame is always the same height from the floor.
As for 1885's single speed in 2014, a post-mortem analysis using video, Ether's Simulation and the original AndyMark orders confirm we were only 10 ft/s after losses, single-speed. There was a mixup somewhere and the end result was us driving about 2ft/s slower than originally planned. In hindsight, 12ft/s single speed was the correct target to get around most defense, and this year we'll have better QA for incoming orders

.