Alright, so let me quickly break down on all methods of calcing Dressrosa's size I have been shown and what might go against them:
Damage’s calc:
-Assumes Beach size is constant
-4 steps
KingTempest’s calc:
-Extrapolates from travel speed that can additionally be uncertain due to the influence of wind and possible other factors
-2/3 steps (depending on whether you count the wind calc itself as a step)
Examination calcs:
1.
-Assumes ships are the same size (reasonable from what we see)
-Low End, as foreground stuff is scaled to background stuff
-3 steps
2.
-Possibly slight High-End as Marine Ship is in foreground
-3 steps
3.
-4 steps, although one is just scaling the height of a human. Otherwise precise.
4.
-scales from something pretty small which is a common source of imprecision
-Assumes Beach width is constant
-2 steps
5.
-To some degree scaled background to foreground, hence high-end
-2 steps
6.
-Scales from something very small, which often causes imprecision
-Background to foreground scaling, hence high-end
-2 steps
7.
-Possibly slight High-End as Marine Ship is in foreground
-4 steps
Now, based on this I find some worse than the others.
I personally don't like the wind calc. Such assumptions are worse than pixel scaling. Yes, the pixel scaling here isn't consistent, but that doesn't mean that using an otherwise worse method is any better.
I think 6 isn't great since it scales from something pretty small.
Aside from that, let's keep in mind high-ends and low-ends, as well as the number of steps, when comparing results:
Damage: 4 steps, 18.17 km
1.: 3 steps, Low End, 3.47 km
2.: 3 steps, Slight High End, 6.7km
3.: 4 steps (1 just scaling human), 5.9 km
4.: 2 steps, 3.3km
5.: 2 steps, high-end, 9.83 km
7. 4 steps, slight high-end, 48.25 km
The two I eliminated had King Tempest's 62.41 km and Method 6's 4.19 km, for the sake of reference.
It should further be noted, that many of those calcs are related. Specifically, Damage's, King Tempest's and Method 4 and 7 all use the same panel in the last step. So the results are technically not completely independent. This might explain why 3 of those 4 calcs result in the highest 3 results in the spread.
In total, it appears to me like taking the double-digit sizes would be neither in the interest of consistency nor safe low-ends.
From the ones remaining after this, 4 and 5 have the least steps, but 5 has also somewhat of an high-end. For 4 one could argue that if the other three that scale that way are considered bad, then this is too. There is also the constant beach width thing, I noted in my list of cons above.
1 is a low-end and hence compatible with higher results. That leaves 2 and 3. 2 is a slight high-end, while 3 has one minor scaling step more. They are also pretty close in value to each other. So I'm very split between them.
I would say that going with 3 is the safest option in that case, for low-ends sake, but if someone says 2 due to less scaling steps I would be ok with that as well.