These tests were done already a month ago (sorry for the delay), but tonight I was finally in the right mood to type in the data, make the graph and do this write-up (I like the testing itself but it always takes some courage for me to digest the numbers and make them into edible graphs :tired: ). Thanks go to MRsDNF for donating me the red and blue XP-E2 (from Cutter, red is bin P3, blue is bin M3), and to telephoneman for the green XP-E2 (from Mouser, bin is R2), they were not easily available to me (in the meantime the red XP-E2 is at intl-outdoor).
As usual, the measurements were done with the leds mounted on copper boards clamped to a big aluminum block, the led wires were directly connected to a power supply for testing. The leds were facing up, with a small OP-reflector on top. Light bounced back from the ceiling was measured with a lux-meter. (These might have been the last leds I have tested this way, for the next one I will try my new integrating sphere :-) )
Note that the output is an out-of-the-reflector measurement, so less light is measured compared to bare led measurements (15%? depending on the reflector, in any case: for all my 3535-size led measurements in the past I have used one and the same reflector).
Note also that I did the usual calculation to convert ceiling bounce lux-numbers into lumens, but for colour leds these lumen numbers are not something to rely on because the cheap lux-meters we use are probably increasingly inaccurate when going farther away from the center of the visible spectrum. so especially the red and blue lumen numbers are doubtful. This means that although the shapes and trends (like at which current the output tops out) in the graphs are as accurate as for white leds, the absolute lumens I trust much less.
On top of all this uncertainty, I find the output of the green led unexpectedly on the low side. In fact, I tested the green die of the XML-colour better than the green XP-E2, even though the XML-colour has an 'older' type striped green die (the xpe2 blue and green dies are dotted, like the white xpe2 leds). I guess this is real, but there's always a small chance that the reflow did not go well (not enough solder under the thermal path), or perhaps the led was a bad one, who knows, I'm not going to redo the test, this will have to do.
Here's the graph, I hope it is useful to some of you :-) . For comparison, I included the red XP-E2 results that I already posted about earlier: