Most manufactures calculate the DR as a theoretical number, based on the SNR of their sensor. They do this because it's a good way to objectively do it based on maths.
All the other methods we argue about involve (rightfully) subjective judgements based on what is acceptable noise. Things like compression, in-camera pre-processing, internal bit depth processing also affect what this number can be. But you can't compare cameras from different brands that use different bit depths in this case other than what you photograph and compare.
These numbers vary because of not only testing methodologies, but subjectively. I laugh when people compare DR on a still frame. The human vision perceives noise differently when it's in motion compared to a still frame. So why compare on a still frame when judging noise is the threshold for DR ?
Arriflex have long understated the DR of their sensor by at least a stop, maybe more. It doesn't matter to me because it's usually "enough" for what I need to do. The best DR claims are the ones I can make for myself when I side by side to a known camera like the Alexa.
In my experience the Ursa Mini is very very close to matching the Alexa for DR, whatever DR they claim.
I even wrote about it here last time I checked it formally.
Since I did this test, I've shot three TV series mixed with Alexa and I've not seen any shots that made me think any less of what I've found here.
I feel so good about this camera and what it can do that I actually CHOSE to shoot Ursa Mini 4.6K over Alexa on the last show I did, "The Warriors".