To an extent, of course, the simulation of reality has to be simplified in a game like this - the trick is getting the right quality and quantity of simplification. As I wrote above, it is beyond any reasonable coding effort to take into account the civil engineering cost of terrain works for traversing any particular way. As to turnpike trusts, do you have any historical data as to the basis on which they charged and whether the charges per mile varied significantly based on the actual cost of construction? The difficulty is that a system in which different sorts of infrastructure (possibly at different times in history) have totally different sorts of access charges will be very confusing to players, and will be very difficult to implement in the code.
I have adopted the Railway Clearing House model because that was a universal system that endured for a very long time and related to what was at the time the most important means of communication. Indeed, nearly the same system was used for canals, too. In "British Canals: an illustrated history" (Charles Hadfield, David & Charles, Newton Abbott, 1966), the author wrote this:
"The revenue of canal companies chiefly arose from the tolls charged at the rate of so many pence per ton mile carried... The charges were lowest for bulk commodities such as coal, culm (slack for lime-burning) and limestone, higher for more valuable bulk cargoes such as iron ore, and higher still for finished goods like iron-castings, and for groceries and general merchandise." (p. 69)
Until 1845, canals were prohibited by Act of Parliament from charging more to use one part of the canal than another, or from discriminating between users (other than by types of cargo carried).
Trying to implement the modern system for UK railways (which relies on a fantastically complex inter-relation of various private companies, state controlled companies, subsidies and payments to the state) would not be feasible as it would require a full simulation of the entire political infrastructure that underpins it (it is not a system that could ever arise by commerce alone). In any event, that system is seriously defective and only exists and is maintained for illegitimate political reasons.
I should note that, for private cars, which do not, of course, run for profit, there is a separate system: see the separate setting private_car_toll_per_km and the comments for it.
As to joint railway companies, these involved actual joint ownership, and that is fundamentally distinct from a system in which one player pays a fee for accessing another player's ways. In the case of joint ownership, both owners would share in the profits as well as the costs of operation, so a system in which the toll is based on the infrastructure cost would not simulate this any better than a toll based on a percentage of revenue in any event.
It would certainly be odd to have implemented a system that is at odds with the well known and standardised RCH system and similar canal system that went before it.
As to whether players have any incentive to allow for sharing of infrastructure, would checking for this not require a precise numerical analysis of whether infrastructure sharing of the type that was actually used in reality (e.g. the trains that ran from London to Edinburgh before 1923 using three different railway companies' lines, albeit usually changing locomotives and crews when they crossed the boundaries) would not be economically viable in the simulation as it is now? That, of course, could only be done after final price balancing had been achieved, which, in turn, is some way off.