Per second tells you the relative proportion of your income that's being lost to fuel. Per truck is only really useful for comparing the merits of advanced resources vs basic resources (which generally won't matter for distant buildings). Per truck would also need the the context of how many trucks will be flying at a time (also highly variable) to be useful. Meanwhile per second covers pretty much all those bases.
You reversed the relative benefits of per second and per truck in every case.
Per truck tells you the proportion of income lost to fuel, by comparing the resource base price to the fuel cost for a single truck. Per second requires multiplication here, resource price * multiplier - per second fuel cost. That's unfeasible.
Per truck is more useful when you're trying to squeeze profits out of a market that you are confident others won't enter soon, and less useful in an already-competitive market. If you're trying to gain short-term profit, per truck tells you the distance and hence startup time thanks to its fuel cost, and per second tells you nothing. The difference in distance matters, even if one of the distances is short.
If you're trying to evaluate secondary long-term profit vs primary long-term profit, it's per second which is useful, since you calculate profit by multiplier * resource price - overall fuel cost. If you had per truck, it would be (multiplier)(resource price - per-unit cost), which results in compounding of the subtraction estimation error. So per second is better here.
Per second is useless in long distance trucking, because to tell if the tile makes any money, you must multiply the base resource by the mining rate before you can subtract. In contrast, per truck is useless in short distance trucking since you know the tile will make money, you just want to know how much, so you'll need to perform the compounding error multiplication.
With that said, your final statement is wrong as well, since going from one format to another is not highly variable, it is multiplied by the mine production. So despite every single one of your statements being wrong, your conclusion still isn't necessarily wrong. Four wrongs make a right. Isn't that how design works?
However, the benefits and costs you considered are not significant. Usually what the player wants to know during tile selection is, "How far away is that tile?" And this is something that fuel cost per truck says, and fuel cost per second doesn't. Distance is what's important, in terms of how long it takes units (buildings, trucks, repair drones) to get somewhere. So per truck is right, and per second is wrong. But remember to divide by 20 to give the unit cost.
Even when comparing tiles of equal resource pips, and per truck and per second give the same number for both tiles, per truck is still better. Because all the tiles around will display the same number, so the user doesn't need to move his mouse precisely over the specific tile to learn the distance. Whereas with per second, it will vary up and down in the neighborhood.
The second important question is what the user expects. For this, it's probably per second, since that's what the building displays. Violating this would require extra UI to clarify, which is unpleasant.
After tile selection, when the building is completed, the per unit cost doesn't matter anymore because the multiplication of prices and multipliers is done for you, so the user looks at the profit/cost numbers on the building popup instead of the resource price and consumption rates. The building shows the per second cost after being constructed, which is correct.