Welcome to BLF Yong!
Any input you can give would be greatly appreciated.
Have a pleasant time at this friendly forum, stevechang!
My dilemma is based on why they did not catch the problems in testing/QC before selling a bunch of them ??? 
1 Thank
I am not authorized to say too much. but as far as I know, in the near future, algorithm employed will be updated, so does the app. the app would contain entry for firmware upgrade. by the way, only white LED light sources are used to calibrate the device.
8 Thanks
LM4 looks accurate when I tested it with Opple house lights.
Lights in house are diffused and not direct like a flashlight. If you put the opple near a window it give you always 100 CRI because is natural light. Try with LEDs and i am sure it is near true measurement but not perfect or not as good as LM3 in this moment
Yeah, I posted some results above and noticed the CCT measure higher and duv higher with some flashlights.
If they only used Opple branded lights for testing then I can see why the LM4 might seem ok. I measured a non-Opple house light and the CCT and duv were off.
1 Thank
Nice, thank you!
There was a statement from Opple that the sensor chip will be replaced in the next batch of the LM4. Can you confirm this?
2 Thanks
That would be good for the future, but not so great for us early adopters. 
1 Thank
Right. Anyway, current beta testers, aka customers, deserve enlightenment, and potential future customers want to know to make their buying decision.
Impossible to cause more damage by an open communication, just the opposite.
3 Thanks
No question about that… and I did not mean to give the impression that I disagreed with your OP on the subject of the sensor. Just lamenting the fact that I feel a bit screwed.
1 Thank
Thats sounds really encouraging!
Thank you.
1 Thank
Yeah, if it was explicitly a test batch and sold at a discount, it would be cheap enough I wouldn’t have really minded (and probably got one and a full production one later), but there was no indication of that. I guess I’d be happy with a partial refund and keep it (and will still probably buy another, to compare if nothing else) if they do change it unless it suddenly became significantly more expensive or something.
2 Thanks
If they need to modify the hardware, they really should offer an RMA for replacement when it is finalized.
3 Thanks
I never heard of chip replacement in this version of Light Master. Next version? perhaps. we are already in discussion with a domestic module/chip supplier, but no final decision has been made yet.
I suggest waiting for app update.
Apologize for inconvenience caused.
4 Thanks
I obtained more reasonable and useful R9 measurements by placing several paper towels (the white kind commonly used in kitchens) in front of the sensor. Here are the results of my measurements:
B35AM 5000K CRI=95.1, R9=84.1 (Emisar D1)
E21A 5000K CRI=94.7, R9=84.8 (Noctigon KR4)
519A 5000K CRI=94.1, R9=84.9 (Convoy T3)
519A 5800K CRI=95.0, R9=79.3 (Emisar D4K)
Latticepower 5800K CRI=91.8, R9=73.2 (Wurkkos TS10)
LH351D 5000K CRI=90.2, R9=63.9 (Sofirn Sp10 Pro)
Surprisingly, the CRI and R9 values for Nichia LEDs did not change significantly after placing the paper towels in front of the sensor, although CCT and duv values did change. The important observation is that the R9 measurements now seem to align better with how my skin appears under these LEDs.
By using eight layers of paper towels, I was able to lower the CCT to the 3500K-4000K range for 5000K-5800K LEDs. It appears that the LM4 overestimates R9 for higher CCT light sources, so I artificially reduced the CCT using paper towels. This method allowed me to obtain R9 values that more accurately represent how my skin appears under the LEDs. Additionally, the R9 values for my other high CCT LEDs now appear more reasonable. When using this method, my SBT90.2 no longer registers R9 values of 100; instead, it either displays a very small number or “–,” possibly indicating a negative R9 value.
One interesting side note: I am aware that the 4000K version of the TS10 exhibits very high R9 values, as noted in Zeroair’s review and other users’ measurments. However, I believe the 5800K version does not have high R9 values, as my skin appears much paler compared to my Nichia LEDs with similar CCT and duv values.
3 Thanks
Thanks a lot for letting us know! 
Has anyone filed a claim on Aliexpress? 
2 Thanks
I returned mine for a full refund on AX, currently in transit in China.
1 Thank
I have returned mine as well