For the remaining 5%, is it only the DNG that is missing or no photo is taken at all? The latter would be expected.
No photo taken at all.
setmode.lua -> results attached; one errorCan you repeat this one? If P mode still fails, please do this:
Miscellaneous stuff -> Debug parameters -> Debug data display [Props], Propcase/Paramsdata page 
and report the value of entry '49' in P mode.
Ran script while in Play mode with no debug. Same error.
Ran script while in Play mode with debug enabled as requested. Lens extends, cam powers off. Lens remains extended. Will not power on (using play button or power button) unless battery pull.
Repeated the "Play mode with debug On" test (3) times in succession, with same behavior each time.
Re-ran Play mode test with no debug turned on. Cam behaves normally, test runs, same error in setmode.log
Ran script while in Rec mode with debug. Pass.
Prop values shown on console: Prop 49 = 33345
Re-ran script in Rec mode with debug. Pass.
Re-ran script in Rec mode with no debug. Pass.
Sorry, I don't seem to be able to get you the props when running the Play mode case.
I see no visible difference between the imagesYou should see differently exposed pictures. If there's no difference that means exposure overrides have failed.
Not sure what to expect re:exposure
Please make sure you run the test in P mode and deactivate 'servo AF' in the Canon menu.
Since the script tries to switch the cam in P mode, it's possible that it failed to do so (as it happened with setmode.lua).
Confirmed that Servo AF = Off in Canon menu
Confirm to run test from Play mode.
ubtest arguments at default.
Comparing images to the "default" image #44 taken by the test;
shoot Tv+2 45 ; is a hair less grainy compared to default
shoot Sv-2 46 ; more grainy than default, pixel blobs visible
shoot ND in 47 ; can't tell apart from #44
shoot ND out 48 ; tiny bit darker in background
I'm ignorant re: photography terms so probably not using the right words. Will seek info to get the proper vocabulary. Also, my eyes are going bad from squinting at this tiny stuff - so it's hard to see subtle differences in the images. After study, the images produced by ubtest *are* different. But not very much to my eye.