Faster, more convenient transportation is what fares are charging for. Non-captured value is more complicated than that.

If the non-captured value is 20% of the captured value, it's highly unlikely that trains will frequently be worth building, but rarely capture enough value. That would require that the true value stay within a very narrow area.

If it's not a monopoly good, and marginal costs are close to average costs, then captured value will only go down as people build more trains, so that value not being captured doesn't prevent trains from being built. If it is a monopoly good (I think it is, but I would appreciate it if some who actually knows tells me), and marginal costs are much lower than average costs, then a significant portion of the value will not be captured. Much more than 20%. It's not entirely unreasonable that the true value is such that trains are rarely built when they should often be built.

That's part of why I asked:

How much government help is necessary for a train to be built?

If the government is subsidizing it by, say, 20%, then the trains are likely worth while. If the government practically has to pay for the infrastructure to get people to operate trains, not so much.

Also, that comment isn't really applicable to what you just posted it as a response to. It would fit better as a response to my last comment. The comment you responded to was just saying that unless the value of trains is orders of magnitude more than the cost, you'd never notice by looking at the economy.