I have been thinking of ways to avoid the issues of the inductor/LED arrangement, which constrains the current pulse width. This morning, I worked through some LTspice simulations using tapped inductors to get different input/output current ratios:
In the schematic, L1 and L2 are magnetically coupled by the K1 coefficient (1) simulating a tapped inductor - this isn't the same as two separate coils. In this case, L1 is smaller than L2, so the input current pulse will be larger than the output. As a consequence, the output pulse will be longer. In this example, the input current pulse reaches a maximum of 120 mA, while the output pulse maximum is 30 mA:
The ratio of the current pulses is determined by tap position. Tapped at the bottom of the coil, this is the original circuit which produces equal input and output current. Moving the tap "upward" makes the input pulse larger while reducing the current in the output pulse and extending the duration.
This could be useful for matching the output to the optimum current to drive a particular LED.
Of course, this presents some difficulty for using off-the-shelf inductors, the majority of which are do not come with taps. Homebrew inductors are easy to make, though, and it might even be possible to add an extra winding on a commercial inductor to create the tap.
I don't know how practical this all is, but it's another possible trick to use.