yeah that's what has me suspicious too, I am comparing using 6 of the 12 12 vs 3 of the cxb3590 both at 150w total
your data suggests efficiency of about 50% for the 1212 @ 25w
and
@CobKits spread suggests around 59%
vs cxb @ about 56%
PCT says for CXB DB bin at 1.4A (48.77W) Tc 25C you will get 9094.6 lm x 3 = 27283 lm @ 186.5 lm/W
citi tool says for 1212 4000k 80 cri at Tc 25C, 700mA (24.2W) you will get 4208 lm x 6 = 25248 lm@ 174 lm/W
my spot tests say that cree@ 1.4A = 435 umol/49.35W = 8.815 umol/W
1212@ 0.7A = 226 umol/23.89W = 9.460 umol/W
mfr data has CXB@7% more efficient under those conditions
real world testing has 1212 @ 7% more efficient under those conditions
i cant call it a fluke when citi beats cree at 3000K, 4000K, 5000K.... how many more chips should i test? heres an old 3000K test. both are 80 cri here. again lets look at PCT vs citi tool
PCT:CB bin: 1.4A: 8395 lm/W @ 48.77W = 172.1 lm/W
citi tool:3000k 80 cri: 4082 lm @ 24.2W = 169 lm/W
advantage: cree by 1.8%
my measurements
CB bin@ 1.4A = 386 umol @ 47.88W = 8.06 umol/W
1212 3000k 80 cri @ 0.7A = 206.6 @ 23.73W = 8.71 umol/W
advantage: citi by 8.1%