According to [1], human ‘fuel’ efficiency on a bike is about 20%. That is way ahead of any simple piston powered steam engine (should be below 10%, citation needed) that you would install in a workshop.
Steam engines are closer to 40% efficient, not 10%.
And they can burn the entire fuel, unlike a person who can only eat the choice parts and must waste large portions. (For example corn - the person can only eat the kernels, the steam engine can productively use the husk and the cob.)
Also, your 20% figure is for energy absorbed after digestion (they check via oxygen consumption), but digestion itself is wasteful - lots of usable fuel is not absorbed when someone eats, and is emitted as waste.
It's really not even close.
It's a problem also for comparing bicycles vs cars - even though they are heavier, cars are surprisingly efficient compared to the entire production process of delivering food to humans (lots of energy is used in the farm).
If I remember my figures correctly 2 people in a car is more efficient than 1 omnivore on a bicycle. And a single person in a car is more efficient (less CO2 emitted) than a carnivore on a bicycle.
(The car comparison depends a lot on what the person eats.)
[1] https://physics.stackexchange.com/questions/46788/how-effici...