Originally Posted by gregf83

That's not exactly a controlled environment for a test. Firstly you don't know if the rider in front of you was putting out constant power. Secondly power reported by a meter is quite stochastic so you'd need to carefully average over a reasonable period of time to get meaningful numbers to do a comparison.

There can be small changes in drivetrain efficiency between different gears which might be measureable with a crank-based powermeter but as I said above the conditions for your test are too variable to measure the small differences that exist.

Suspect you were just seeing coincidental changes in power.