You are right: AL's "differences" test produces different results than Ozone's.
What I did was export a portion of a project twice, first with Ozone's artifact-monitor on and then with AL's artifact-monitor (both set to 320 kb/s). I exported them as 32-bit waves to avoid any differences from encoding or bit depth changes or dither. I then imported the two files into a new SONAR project and level-matched them. I then used exclusive-solo to bounce between them.
To my surprise, they were audibly different from one another. Of course, they did not null, either. It does not appear that one is superior to the other, just different. Still, it raises some interesting questions about how two MP3 encoders could produce different results, given that they are both presumably using the same patented algorithms.