> Incandescent lights used substantial chunks of your pedaling power.

Not solely caused by the lights as they are about ~5W anyway (edit: the old one in my box of bike parts says 6V/3W on the metal). The wheel dynamo's are insanely inefficient and get hot everywhere which were the primary ones used with most incandecant lights.

hub dynamos are not particularly efficient. it's very difficult to make an electric machine that is efficient at the (very low speeds, relatively) that a bike wheel turns at. 60% sounds about right. However, safety lights use much less than 5W, and a modest but very useable headlight about 3W, so your figures are otherwise pretty close.

100W is pretty high for an untrained person. Back when I was cycling regularly I could hit those numbers but not every time. I took my first ride of the season and averaged a whooping 63 watts. Stealing 5 watts from that it taking away 8% of my energy. If you account for efficiency loss in the dynamic it's probably closer to 16% of my output. That's a huge chunk

The dynamo wouldn't necessarily need to provide full power. Say your commute is 1 hour. The dynamo can provide 3W and the light draws 5W. Your morning commute would provide 3Wh to the battery. You don't use the light so all power charges the battery. During the evening, you now have at least 3Wh of battery alongside 3Wh of pedal power so you'd have 1Wh of battery left by the time you get home. A large enough battery could store hours of surplus light just based on your normal daily commute. If not battery power is available and it's night time, the light could be dimmed or you could charge the battery from USB. If the light is off, then load on your muscles from the dynamo would decrease as the battery charges to full.