It's impossible without actually measuring, which I'll come back to. But are you aware that the difference high to low in the various estimates is less than 1% (0.8%).
In your shoes, I'd toss out the 2125 number and assume it's somewhere closer to the 3 lower numbers which vary high to low by less than 1/4% so I'd go with 2108, and probably be within 1/8% of the true number. If you averaged all four you'd still be at 2112 so pick something between 2108 and 2112 and accept that you can't be more accurate than that.
If you want to measure you can do a rollout. You'll need a plumb bob, (weight on a string), a tape measure, and a piece of chalk. Start with the valve exactly at the top or bottom, verified by the plumb lining up the axle and valve. Mark a line on the pavement. Roll the bike in a straight line exactly one revolution (check with plumb line) and mark a second line. Measure the rollout (circumference). If you have a friend to work with, it'll be more accurate of you sit on the bike to compress the tire during the rollout measurement.
Now that you think you have the computer calibrated to within 1/2% accuracy, here's the bad news. Front wheel computers will always read high. That's because the front wheel rolls farther than the back wheel, always tracking a wider arc on all curves. Since nobody rides a perfectly straight, and roads have curves anyway, this can add up. If you want ultra accuracy you'll mount the sensor on the rear wheel, which tracks a straighter line.
In short, don't go crazy, use 2108, or something slightly larger to compensate for the tracking error and be secure that you're accurate enough, within 50' over a mile.
__________________
FB
Chain-L site
An ounce of diagnosis is worth a pound of cure.
Just because I'm tired of arguing, doesn't mean you're right.
“One accurate measurement is worth a thousand expert opinions” - Adm Grace Murray Hopper - USN
WARNING, I'm from New York. Thin skinned people should maintain safe distance.