I thought the paper (or maybe a different one) specificially compared efficiency of riders at their self selected cadence only, plotting the trend based on people using different selected cadences. I'm not certain though. I'll have to look again. I also don't pay any attention to cadence actually other than what feels right or with a little push up or down for some training or situational reason. I'm pretty sure if heat puts pressure on the situation that the "feels right" approach will react well to that, but I never took notice.