It is more efficient, per se...
B-29 crews in World War 2 found that when unloaded, their four-engine
aircraft could fly farther on three running engines and one engine
stopped (and its prop feathered to minimize drag) than they could with
all four turning. Why? Because cruising speed with just three engines
required opening their throttles more, reducing pumping loss.
Engineering schools run well-publicized fuel economy contests that are
now won by fabulous numbers such as 1200 mpg. This is achieved by very
lightweight, low-drag shapes that are accelerated by tiny engines
running mid- to full throttle. Then the engine is disconnected and
stopped while the vehicle coasts down to some chosen low speed, at
which point the tiny engine is again started and accelerates the
vehicle again. The engine’s small size, need to run at large throttle
opening, and intermittent operation minimize both pumping loss and
mechanical friction. Mileage fanatics, driving instrumented Honda
Insight two-seater cars in this same “pulse-and-glide” style, have
been able to get close to 100mpg on long trips.
Of course, pumping loss isn’t the whole story in the above cases.
Cylinder deactivation also saves some piston ring and bearing loss in
the deactivated cylinders, and clearly, three B-29 engines have less
total mechanical friction than four.
Charles Lindberg used a similar technique to significantly increase
the range of the P-38 fighters in the Pacific. He showed the pilots
that reducing rpm, leaning the mixture and INCREASING the manifold
pressure drastically reduced fuel consumption.
I suspect that this is not implemented in consumer cars because it is not very "easy" to use and is not as economical for manufacturers to provide.
As you can see from the above examples, to really be efficient this method favors having multiple smaller engines all running at mid to high throttle over having one single powerful engine running at a range of throttle levels.
The costs for manufacturers to provide a vehicle with multiple engines to efficiently account for each speed range would be enormous and is probably not worth the savings imparted to the customers through fuel savings. Not to mention the increase in maintenance by a factor of at least the number of engines in the vehicle.
But what if we said screw the multiple engines, lets just try to make the one engine more fuel efficient by modifying the gearing?
Then the car, at any given speed, will have one optimal gear for acceleration and another optimal gear for cruising for fuel efficiency.
I think on average this method might gain the average person about 1 mpg, maybe 2, at the cost of real-time throttle responsiveness. I think if you try to balance the scale between throttle response vs 1-2 mpg, the benefits of the mpg over millions of people is definitely there (1-2 million miles or roughly 50k-100k gallons of gas or roughly $200k-$400k per million people), but is it really significant enough to outweigh the few accidents that may have been prevented by having better throttle response (and the added "comfort" of having some throttle cushion during normal driving)? It would really depend on the statistics and analysis therein, but I would guess probably not.
Additionally, especially for automatics, this method would have to be heavily controlled by the computer to be acceptable for use by everyday consumers and would cost that much more money in R&D, testing, and production as another feature they provide. If a feature is provided but does not work correctly or smoothly enough then it damages the brand and might cause direct damage through warranty work or recalls.
Honestly, I think you have a better handle on pumping loss and frictional loss in gasoline IC engines than the other responders. Might have had better responses in Engineering SE.