This may be obvious, but...
Keep in mind that the less gas you need for a fill-up, the BIGGER impact even a 1/10 gallon difference will make in your MPG calculation.
Example - 100.0 miles driven on a tank, then fill-up:
First time it takes 2.000 gallons = 50 MPG.
Next time it takes 2.100 gallons = 47.62 MPG.
Did my MPG really go down, or was it a slight difference in when the pump clicked off (or otherwise in how I filled 'er up)? One may never know, but let's presume it WAS differences at the pump that introduced this 2.38 MPG difference.
Filling a "closer to empty" tank is always a little more accurate way to compute a tank's average, than is "topping up" and calculating that average.
Example - 500.0 miles driven on a tank, then fill-up:
First time it takes 10.000 galllons = 50 MPG.
Next time it takes 10.100 gallons = 49.5 MPG.
Whatever the gas-pump variable was has now only influenced the MPG by 0.5 MPG.
IF the tanks were filled identically each time, it would make no difference (except for rounding the miles/km to the nearest tenth). But that's difficult to do.
Bottom line is larger numbers work better for averaging than do small ones. For MPG, this is due mainly to differences in "how full did you really get it THIS time vs. LAST time". Also the accuracy of the pump's display of gallons/litres dispensed, and the accuracy of the odometer come into play. The odometer's accuracy (or lack thereof) is pretty consistent.
You can decide to always fill up using the same pump to reduce this variation.
Even the odometer may report distance inconsistently over time due to the wear of the tires.