How do you know the computer isn't correct and the sensitivity of different fuel dispenser shutoffs is making your hand calculation incorrect??
When I top off the tank, I can usually add over another gallon to the tank after the initial automatic stop of the pump. Some pumps will let you add a lot more than others. This translates to up to around a 5% difference in fuel added which calculates out to around a full 1 MPG variation depending on where the pump stops your fill. With that in mind, the accuracy of a "hand calculation" of any given tank can be off by a full MPG. It is no wonder that the numbers do not align with the computer, but occasionally do. It is the average hand calculation of many fill ups that will show the accurate number. Unfortunately, when you compare the average of many hand calcs to the computer, they STILL won't line up because I believe the computer averages the MPG over a rolling 300 mile span....which is roughly 1 tank. With that said, it seems that it may be impossible to accurately compare hand calculated MPG to the average MPG as shown by the cluster. Lame eh?
First, I mostly fill up at the same station. Maybe that matters. Maybe it doesn't. I do not use the same exact pump every time.
Second, I fill until it clicks off, wait 5 - 10 seconds, then pump again at about half "speed" until it clicks off again. It generally only accepts another 0.2 gallons or so. The way I see it, it doesn't matter what the initial cutoff is. If you keep putting gas in until it is clicking off immediately, then I reckon it is full - to the same approximate level every time.
Third, I generally run the tank to near-empty, so I'm putting in more than 25 gallons at a time. The volume and number of miles involved serve to reduce the margin of error. When I talk about how "off" the onboard computer is, I don't even consider the differences between onboard and hand-calculated for fill-ups where I put in 10 gallons or less. The margin of error on those is too high to bother considering.
Also, I reset the tripmeter and the Fuel Economy screen every time I fill up. Unless the way FE is calculated is totally retarded, that SHOULD mean the FE screen and my hand calculation on an individual tank should always match (within the limits of accuracy of the way I fill the tank).
Everything you said still doesn't account for differences of 2 MPG between the computer and hand calculation (again, over fill-ups from a near-empty tank).
If the difference in numbers was because of big differences in the actual fuel level I achieve on each fill-up, then the numbers should be distributed pretty evenly between the computer being optimistic (i.e. reading high) and pessimistic (reading low). But, they are not. The computer almost always says I got better mileage than I really did.
My overall average (of 14.4MPG or so) is the total number of gallons I have put in over all time divided by the total number of miles I have driven. So, the margin of error for that calculation is VERY small (presuming the odometer is accurate, of course).
The bottom line: Hand calculating is pretty darn accurate when you do two things: Limit your observations to fill-ups from near-empty, and consider the data over a number of fill-ups. And, unless they have been doing a reliable, reproducible hand calculation every time and comparing that to the onboard (which gets reset every time), anyone claiming that the onboard computer is reliable and accurate is really just spouting wishful thinking. Maybe theirs is. Maybe it isn't. But, if they haven't checked it a bunch of times, over a bunch of miles, there is no way to know whether it is or not.