Hello Toykeeper,
I have also have the D25A N219 and I meter mine at:
Eneloop (~1.3ish V): ML on - 0.15, 6, 60, 100 and ML off - 2.5, 13, 60, 100.
I would use ETs “6x and 2x” footnote for ML off so ~3 and 14 should be spec for the N219. The 4/20 lumens you use is applicable to the more efficient CREE emitters, IMHO. As you can see, except for ML, I match-up to spec quite closely.
Regarding moonlight mode, I personally like brighter ones, and was disappointed that mine was so low. But then I ran a side-by-side moonlight mode runtime/efficiency test and noticed that the D25A’s issue was poor regulation on ML - it is highly voltage sensitive. On a AAAA (from a 9V) it started out @1.6v and over 0.5 lms and fell to 1/4th output as voltage dwindled. My ’12 XML also shifts on V, but only by half. In contrast, the Quarks and SC52 have stone flat regulation on ML. Here’s a photo of that test, somewhere in the middle of the run:
Since I prefer brighter ML modes (my most often used mode, and represents a my general purpose low with dark-adapted eyes) I now only run an L91 in my N219 which gives me a ~ 0.6 lm moonlight, slightly dimmer/floodier than the D40A of the top picture. My other favorite ML is the 0.3 lm from an XML Quark.
I can not comment on how much is due to sample variation, or what voltage you were testing the light with, but I can understand how the D25A N219 appears to be so bright calibrating from an ZL sub-lumen scale (mine is also ~10x brighter - 0.6 vs 0.06 when spec is 0.5 vs 0.34). I calibrate to 47 lumens, which quite closely tracks ET (and most other brands I collect from), and it is the SC52 that I find miles off, especially on all the low modes which are off by multiples…. but I agree with the ~40% ish differential you mention for the higher modes.
I know we’ve had this discussion before, and that you like Selfbuilt and ZL’s lumen scale, but I prefer ti-force’s and these two reviewers will differ by ~70-80 lms on the same 300 lumen light. And interestingly here’s what both have to say about their accuracy….
As one more example of how far off the SC52 moonlight specs are, here’s the three moonlight levels (0.34, 0.06, 0.01) shown between the 0.09 lm fireflies of the (early gen) T10 and L10 (which both SB and I find to be ~0.02 lms). Another member or three on CPF have confirmed similar, and we’ve previously compared beam shots, so I don’t think its sample variation.