We can identify scenarios where recuperation could help a lot:
- urban stop and go as discussed in this (German language) article: they estimate that starting from a stop happens with a total power input of 200 - 300 W (depending on whether there's e.g. the groceries to transport etc.) for the first few seconds.
- tours with many not-too-long ascends and not-too-steep descends. Steep descends mean that only a small fraction of the potential energy can be harvested.
The article argues that someone who is not a fit biker (but e.g. an elderly person doing their shopping) may easily have the muscle power to go at an acceptable speed in the flat once at speed, but they may have difficulties producing the power output to accelerate the bike/hold the speed uphill that allows a safe and stable operation of the bike.
In fact, they basically argue for an electrically assisted system that helps only at low speed - while not hindering the normal muscle output above that speed. The idea is to get rid of peak power, so the biker can get along with their own continuous lower power output.
Hills need some more power stored than acceleration after a stop, but such a system could get away with very small batteries (see below).
The idea here is very different from sportive e-biking: the assistance is meant solely to help people avoid people becoming so slow that the bike becomes unstable.
The competitor of recuperation is a bigger battery, and IMHO this is where it is not considerd worth while.
A glance through the internet tells me that one can get 1300 kWs (360 Wh) in maybe 5 kg of battery.
Going with @cmaster's back-of-the-envelope calculations, that translates either to 500x accelerating from stand to cruising speed or 1300 m of elevation gain.
With a recuperative system as outlined above, we could get away with a small battery of < 1 kg. Without recuperation, 1 kg of battery using the electical assistance only as outlined above would still give us the equivalent of 250 m elevation gain or 100 starts. Plenty for every-day use (and the scenarion is not to stop pedaling, so this would cut 500 m elevation gain in half or so). But 1 kg battery is still small compared to the weight of the drive. (Yes, with the target clientele of the scenario above, a 1 kg battery instead of a 5 kg may be an argument...)
And 5 kg battery can be marketed as assisting not only for acceleration and getting us to go uphill at maybe 8 or 9 km/h but at getting us to an acceptable overall speed for so many km. And it can be marketed in line with sportive e-bike - which the outlined system would not be.
And incidentally, going assisted all the time will hide the resistance due to big low-pressure (or even knobby) tires and a not-so-very-efficient drive train.