Here you go...
I figure, if you had a mind to, you could add regulation by running this dude in burst mode, by stopping the 555 via its RESETbar input.
Diode D7 prevents some really weird behavior from back conduction. If there were a need to tighten up the efficiency, I'd replace it with another P-channel, body in the opposite direction from MP1 (for lack of finding a suitable transistor, modeled as a 0.3 Ohm device extrapolated from my vast CMOS device library :-). Then I'd use a comparator to make the drop like a near ideal diode as in...
My experiments with using an auto-transformer would indicate that the current capability of the hub dynamo (at least as presently modeled... I'm suspicious of its accuracy)) falls off rapidly, almost acting like a (near) constant power device.
I'd ask one of the local "pansies in spandex" to make experiments with their hub dynamos, but I'm their enemy for suggesting publicly that they should be required to be licensed, and be fined when they go riding three-abreast down the middle of a thoroughfare, rather than single-file in the bicycle lane :-) ...Jim Thompson