Results of the Yuji high CRI 5mm led tests.
While doing the output/current tests of the 3200K and 5600K version it became very clear that these are different leds, not just a different phosfor. Then I took a closer look and the build appeared slightly different too. At first that made no sense to me, the difference should only be the phosfor. But then I remembered that the 3200K and 5600K batch were purchased with 2 years in between, 3200K in January 2015, 5600K in May 2017, so in between Yuji had changed the design (for the better, as seen in the test) So most likely the 3200K leds in the new Sofirn behave like the 5600K led in this test, apart from the tint.
For the test I thought it would be fair not to cool the leds as well as possible but simply solder them on a common FR4 board, so the heatsinking would not be too different from how the led is mounted inside the flashlight.
Here are the output/current tests:
Directly after the test the leds and led wires felt luke-warm to my lips, nowhere near hot.
This chart already shows that the chosen 25mA is a very conservative current, at 50% or less of the maximum possible output and 30% of maximum current, while most of my flashlight builds are 80% of max output or more.
But to get a hint at longterm stability I soldered the two leds in series and did a 24-hour runtime test at 50mA (so double what the leds will get in the flashlight). Here are some voltage and output numbers of the combined leds in series:
0 hrs 6.43V 29.9lm
5 min 6.37V 28.2lm
30min 6.35V 27.6lm
2 hrs 6.37V 28.0lm
3.5hr 6.38V 28.0lm
10hrs 6.39V 28.3lm
16hrs 6.35V 27.5lm
24hrs 6.38V 27.7lm
So, nothing exciting is happening here, the leds are fine and show no signs of fading. So unless led death is a sudden event without performance warning, also 50mA is a fine current for these leds.
AFAIC, the suggested 25mA drive current for the flashlight is a nice careful current.