I believe my understanding of your test is correct, but your understanding of IR's test is not. Please correct me if I'm wrong.
If you read their description, they don't add actual focus time. The lens is already in focus. It's the time for the body to realize it is in focus:
Quote: To minimize the effect of different lens' focusing speed, we test AF-active shutter lag with the lens already set to the correct focal distance. Being an older design with a non-ultrasonic motor, it wouldn't be the fastest at slewing from one focus setting to another, but that's exactly the reason we measure focus determination speed, which is primarily a function of the camera body, vs focus adjustment speed, which is primarily a function of the lens.
They are all shot taking lag with slightly different situations.
Regardless, the relationship between prefocus lag or AF mode and MF should be consistent between the tests. That leads me to believe firmware changed something, as we're not looking at absolute time differences.