Possible Bug in Measurement "Check Levels" button

New user here. As I learn about the software, I've been trying to keep tabs on behaviors that seem to interfere with the learning process.

One issue: I have not been able to get "Check Levels" to consistently report an output db that matches the calibrated level of 75db. Given that REW does report the correct SPL in other contexts, this doesn't seem like a necessary or intended behavior.

Steps:

1) Run through the "Check Levels" procedure to get -12 dbFS signal level to produce a measured 75dB volume level, using a main speaker test signal, and reading from an external SPL meter.
2) Calibrate the SPL level, confirming that the main speaker test signal still produces 75 db on the external meter, and adjusting the calibration control to match
3) Use the REW signal generator and the REW SPL meter to confirm that both REW and the external SPL meter now agree with each other at around ~75 db
4) Open the measurements window, enable "Use loopback as timing reference" (wiring the system appropriately), keep an eye on your external SPL meter, and click Check Levels

Result: The external SPL meter will still read around 75 db during the test period, but Check Levels will come back and report 80db (sometimes 81db), along with ".
Expect: It should report the same 75db SPL observed externally.



Note that if I turn off "Use loopback as timing reference", then clicking Check Levels will come back with the expected 75db. I don't know if it's just reporting the louder of the two signals when the timing signal is enabled, or if it's adding them together to simplify the level check somehow, but it's confusing behavior. Does it even make sense to measure a loopback signal in terms of SPL considering it is not producing an actual sound?

Sorry, I should have included the whole thing the first time, I just wanted to make sure the problem was clear.

Here it is with no timing reference:




And again with timing reference (and the unexpected elevation of the levels):



The change from 75 dB to 81 dB suggests that the ref and output signals are being added together somehow. It also looks like the dBFS and SPL numbers are changing the same amount (+6 in both cases), which may indicate the SPL number is using a suspicious formula for addition (I might generally expect a dBFS +7 change should be equivalent to a bit over +3dB, not +6dB), but that's kind of ancillary.

If I unplug the microphone, a condition that Check Levels should really shine at discovering, I get arguably the most important symptom, a false success:



Finally, with Microphone unplugged and NO loopback timing reference, the result is again as expected (a "Level Low" failure, on account of no mic):



Let me know if you want to see anything else.

Thanks for all the work you've put into this impressive program!

Direct Monitor is off.. when it's on there's a feedback loop, as you might expect (I've experienced this), just for extra confirmation.

Are these behaviors unexpected, then?

I think I did still find the problem in Focusrite Controls: in the device settings "Combine Input 1 and 2" were checked. Looks like this was having an impact on the Levels as well; they were attenuated in the mixing process (probably to avoid clipping).

I don't know when that got enabled, but I must have done it as a diagnostic and forgotten about it. Thanks for the clue, it looks like I'll have to recreate every measurement I've already taken.