ISTR a study of pro cyclists that showed that work efficiency was maximized at lower cadences -- but I always thought this was an argument that while racing, pros don't tend to maximize efficiency, they're more interested in racing results no matter how inefficient it may be. Race cars get much lower mileage than economy cars, but they're usually not trying to maximize mileage (race cars have their own efficiency and fuel consumption goals, but those goals aren't to maximize overall economy, they're to get acceptable economy compared to the other cars they're competing against).

I've looked at the cadence and power (and crank torque) of several pro races. Although their *average* cadence is relatively high, they tend *not* to stay within a narrow cadence band: they're not automatons, their cadence band spans the range from slow to very fast, depending on the race situation. That said, the range in power is "wider" than the range in cadence, so they *tend* to vary their crank torque to modulate power (that is, they vary both cadence and crank torque to produce the power they want, but they vary crank torque more than they vary cadence).

When I've looked at data from me and some of my friends (and we're definitely not pros) the overall pattern of wide cadence range, not being automatons, and modulation of power more by crank torque than cadence, seem to describe us, too.