Is the air conditioner more efficient than suffering drag from open windows? Each has been commonly recommended at different times, and the correct answer depends on your compressor and body shape. The optimal cruising speed might be anywhere from 40 to 60 MPH, depending on your engine torque, transmission, and tires...
So you're trying to decide if you're better off with the A/C on or the windows down and you trust the short-term estimated MPG over the measurement you could do yourself? Fill up the car and turn on the A/C, drive an hour. Refill the car and record how much gas you put in. Turn off the A/C, roll down the windows, drive another hours. Refill and record. Now you can answer for yourself.
I get the logical appeal of the immediate feedback MPG meter, but I'm not clear how accurate these things are (vs meters on the gas pump, which are regulated to be within a tight spec). I'm also not clear how useful the MPG right now is vs the MPG over the past hour. I've seen only one car that let me reset the MPG meter and measure for an arbitrary period of time that I control, and I still have no idea how accurate it was. (Yes, I'm sure there is more than one car with this feature. I've only seen it once.)
Again, I just generally don't buy that instant (actually averaged over some amount of time that you don't control, with some accuracy that is likely not guaranteed) MPG estimates are all that useful. Are you dedicated enough to efficiency that you'll drive without A/C in the summer to see if it's better but you somehow don't care enough to bother recording actual gas usage? Who are these people willing to drive without A/C to save a MPG but too lazy to confirm that it really works? Or realistically is this one of those things where you turn off the A/C, glance at the MPG estimate and see it went up from 20 to 22, and assume that you've got a definitive answer despite not knowing anything about the margin of error for the MPG estimate?
On one hand, instantaneous mpg is seriously misleading. To me, accelerating efficiently means maximizing the amount of acceleration I get per unit of fuel—doing that does NOT maximize instantaneous mpg, but ideally finds the optimal point on the fuel-consumption-rate*time graph that you actually care about.
On the other hand, real-time feedback is probably better for the not-exactly-scientifically-rigorous way people learn, and it can be useful for noticing unexpected behavior not covered by intentional experiments.
A realtime unit-less fuel efficiency gauge in a U-Haul clued me into non-ideal behavior from the automatic transmission climbing hills. Feedback from this dumb gauge caused me to adjust my driving, which meant only my first tank of gas lasted way less time than expected.
