Yeah, the basic circuit is dumb and sucks. It's an easily-built curiosity, but not practical or efficient in any meaningful way.
Note that 600kHz is
way too high for a #26 core -- the #8 should be okayish though. These are often built with ferrite beads, which is just all kinds of wrong, but, there you have it.
I made this one about five years ago,

You can basically read the circuit off the layout. Selectable pull-up (infty, 1k, 100R) for adjustable light level. 22uF ceramic for main bypass, 1uF I think for base bypass. I think the primary is 2 x 28 AWG in parallel. The secondary I think is 2:3 or 1:2 ratio, so stepping up slightly. Core is #8 (yellow/red). There is no added resistance between the base bypass cap, secondary, and base pin. This can be added to soften switching, but it almost always only worsens performance. Transistor is a PBSS303NX or similar, low-Vce(sat) type, rated for several amperes.
Input current is around 1A, and output is around 1W (4V pulsed, white LED). Forget what switching frequency is, maybe 200 or 500kHz?
I evaluated efficiency by measuring input current, and measuring output voltage and current using a schottky diode. In that configuration I got 67%, not bad at all.
If I ever made a bigger one, none of this would be necessary; for example, my 10W flashlight is build around an 18650 cell, so little conversion is necessary. As it happens, I chose to use series-connected LEDs, and an average-current-mode boost converter:

Tim