The test is in Top Velo April 2011. It re enacts a test they first did in 2004. A rider did intervals of four minutes. Cadence was fixed at 100rpm and HR at 160bpm. There was an 8 minute rest between intervals followed by a two minute warm up. The order of the pedals was random on changed throughout. The height of the saddle was adjusted and the alignment of the cleat to the pedal axle was also corrected. All data was recorded using a SRAM S975 Quarq.
Repeated how many times?
There's your rub...
If this wasn't repeated in alternating order (the list forward, backward and alternating from the middle out) several times where the watt diffs were repeated regardless of the pedals place in the order, the test just doesn't hold water... The bike would have needed to be refitted for stack height each time as well as making sure the cleat placement and alignment were perfect each time. I've seen larger wattage difs with no changes in equipment but slight aero tweaks, for same heart rate in pro athletes doing fewer runs...
The difs are also, for the most part, inside most power meters margin of error during same calibration.
If they did repeat the tests and fitting and the watt difs from one pedal to the next (which would be more important than producing the same total peak wattage) were the same each time, its a fair job of holding up.
If they didn't shuffle and repeat and fit properly, there's virtually no usable info.