My LM4 readings are from the reverse-engineered app v3.3.1. I tested a few measurements with the Windows app, and the DUV numbers are similar to my previous results. I have not noticed much difference in DUV measurements across different versions.
For my purposes, the 3.3.1 app works better than the desktop version because it seems to better capture my perception of red light in the spectrum. (See here for how I perceive red light in lights). To be specific, R9 measurements tends to be over estimated in the desktop version for the lights I have.
Here are my R9 measurements from LM4 for desktop and android 3.3.1
TS10 6000K, [Windows] 82.1, [v.3.3.1] 58.9
LH351D 5000K, [Windows] 60.5, [v.3.3.1] 30.0
519A 5000K, [Windows] 86.6 , [v.3.3.1] 91.3
E21A 5000K, [Windows] 82.5, [v.3.3.1] 93.0
For me TS10 5000K has far less pronounced red spectrum than either E21A or 519A, but the R9 numbers are about the same in the Windows app, while there are significant gaps seen in the v.3.3.1.