Same window
Miles and gallons from the same days.
The example uses 172,800 miles and 25,890 gallons burned, which is about 6.6 MPG. Card purchases were 26,940 gallons. Using the card total would have told a different story and a worse one.
I will compute yours the same way only when the burn figure deserves it. If tank data is junk, I will not publish an MPG that depends on it.
It is for spotting tractors and lanes that burn more than their twins, after theft and mixups are out. It is not for catching a stolen fill. The match does that.
Spec sheets and OEM targets are your business. I will not grade a fleet against a brochure.
What I look for
Miles and gallons from the same days.
Units whose sensors or burn method you accept.
Stolen gallons are not an efficiency problem.
Short answers
I can stop bad gallons from distorting it, and I can show idle and habits afterward. I do not retune engines.
Then I will not invent MPG from the card. We stay with the match.
Next step
15 minutes. Tell me whether anyone trusts the current number.