I agree, it doesn't seem logical. However the speedbonuses and maintenance costs were all calculated so that there is a broadly equivalent profit per passenger per kilometre between all vehicles, as long as they are in the right era.
I will point out that this is not actually quite the correct thing to do! In many eras, one mode of transportation was quite dominant with others only being used in specialized situations.
It's probably not worth messing with it for simutrans-standard, because only simutrans-experimental has the features needed to do the modelling correctly. However,
(1) Prior to the invention of steam trains, transport by ship was *much* more profitable per passenger-km (or goods-unit-km) than transport by road. (After that, the ships were too slow... before that, the ships were often *faster*, but they were cheaper to run even when they weren't faster.) If time isn't an issue, barges are still very cheap to operate and therefore give excellent profits.
(2) Transport by train has always been more profitable per passenger-km (or goods-unit-km) than transport by road -- provided the train is carrying a large volume, and the road has to be built and paid for. The advantages of road transport are in low-volume applications, and because the capital expenses are lower.
(3) Trams are more profitable per passenger-km than buses -- but have much larger capital costs.
So
(1) there should be a capital cost/operating cost tradeoff. The methods of transport with high capital costs should have higher operating profits and vice versa. Perhaps your spreadsheet already includes this; to do it correctly you have to annualize the capital cost over the vehicle lifetime, but only in simutrans-experimental do the vehicles really have lifetimes. Otherwise, you have to work out an expected return on capital per passenger-km; a twice-as-expensive bus should give you, perhaps, twice as much money per passenger-km. This is actually part of the ship advantage; when they're fast enough and the water's already there, ships are exceptionally profitable, but the capital costs of canals are outrageous.
(2) There should be a volume bonus. The operating profit per passenger-km should be just slightly higher when you use larger-capacity vehicles -- provided the vehicle is full and running at full speed, of course. This makes people "right-size" their bus, train, and ship fleets, as in the real world. (However, since waiting time 'doesn't count' for speed in simutrans-standard, people can just leave a large bus and wait for a long time for it to fill up -- so this balancing only works in simutrans-experimental.) Running as many small, cheap vehicles as you can until you hit congestion is often highly profitable in simutrans-standard, and that's unrealistic particularly for goods which aren't time-sensitive. You could also keep the operating profit constant and make the capital cost per passenger (or unit good) lower for higher-volume vehicles, which has similar economic results, but it's more realistic for the operating cost to be lower.
(3) There should be a slight bias towards the historically correct method of transport and away from "outrageous" choices, although if (1) and (2) are implemented and the simutrans-experimental revenue model is used, this will mostly, though not entirely, happen on its own. The most oversized super-giant buses should be less profitable and more expensive to run than similarly-sized articulated trams (people will still use those buses occasionally to avoid building a tram network); trains should be the most profitable long-distance transport in the late 19th century (but people will still use buses for low-volume long-distance transport), while ships should be the most profitable in the early 19th century and before (people will still use land transport because canals are very expensive to build, but they won't use land transport along the coastline).