This is just a quick note to say I built a mk2 of this charger, largely based on the original circuit here, with the overvoltage protection (TVS/zener triggered thyristor on the DC side) suggested by DoctorRad here (and also the schottky diode on the ground pin of the regulator, to send it slightly past 5V): viewtopic.php?f=5&t=86041&p=802691#p802691
I packaged the whole thing in a plastic tube (I also reused the USB socket from a car USB charger which I gutted). Was almost going to fit a switch too, but decided to keep it clean.
And it seems to work. I can get my LG G2 charging at roughly 1% per km. It's a big 3000mAh battery, so that's not bad at all. All the actual testing of the overvoltage aspect was indoors, so I can't vouch for it in all circumstances, but nothing has gone wrong.
I also bought a Tecknet iEP360 6000mAh power bank and tried charging that (largely as I want to avoid charging the phone directly). It seemed stuck on two lights out of four for ages before I worked out that if I start charging from a complete stop, the power bank shows one light, which either means it's not happy with the initial voltage or current (too low), or is otherwise only accepting a limited charge rate. But if I unplug the charger and replug it at 8km/h or more, such that the correct voltage and decent current is immediately available, as would be the case for a mains charger, the normal charging flashing sequence lights up. A little inconvenient (I should have fitted that switch!) but knowing this I have now been able to charge at roughly one light (20-25%) every 60km. I believe quirks like this on the receiving device are responsible for many people having issues charging certain devices. I know the Bright Bike (now Sinewave) Revolution charger has a programmed 10 second delay before it delivers power, possibly to counter this very problem.