Different OS? Different color settings?
Image comparison checkpoints are NOTORIOUS for being imprecise. When doing the checkpoint, the log should show a "mask" showing, in red, the different pixels. Anything stand out? Is there a "date" or "time" on screen that is different each time? Lots of things could cause a difference... even if everything is exact, there's always a little bit that's off...
Yes, its hard to make image comparison robust and reliable. But still achievable.
Some hints:
- create region reference image on the same computer the test run
- use mask
- use pixel and color tolerance
