The high road calculation comes from mergers. Figure it this way:

• During setup 5 companies are floated, each of which builds 1 free route
• There are 7 possible Expand actions per round
• In the first 3 rounds each Expand can place 1 route
• In the last 4 rounds companies can double expand (place 2 routes), but this is modestly rare due to cost
• There are 5 secondary companies that can float during the last 4 rounds
• Each of the secondaries, if floated, can build 1 free route
• Companies can merge. Not infrequently the game will end with only one company (the putative British Rail) left in the game
• When companies merge the route markers from the acquired companies are replaced with route markers of the acquiring company (ie the board goes increasingly monochrome).

Ergo by the time the game ends there may be 5 + (3 * 7) + (4 * 7 * 1.5) + 5 == 73 route markers on the board, all of which are the same colour — and you don’t know which colour that will be as of the start of the game! Aside: This assumes a 50% double build rate, which is high, so I took 70 as a reasonable rate. In practice the total number of routes on the board will be smaller. All 5 secondary companies won’t always float, in some rounds all the Expands won’t all be used, the game will regularly end in the 5th or 6th round instead of the 7th, etc — all leading to less total routes placed during the game. But the game manifest needs to account for the pessimal case — not having enough bits in the box to play a predictable game-case is Not Good.

If I do release the game under a CC license I don’t expect people to use wooden route markers. I expect they’ll use crayons, grease pencils or wet erase pens. In that case the bit supply simply won’t matter. However I wouldn’t publish the game that way. Much as I like pen-based routes I don’t consider it a viable commercial product characteristic.

Oh, and 700 roads? How many can possibly be used at once? In one on my games, I had spaces for 200 roads and no matter what I did, I simply could not put it on a board that small.

If you had to, you could probably get away with squeezing the land mass a little bit so it’s not quite exactly accurate, but no one would care — kind of like how US maps sometimes have a straight top border and sometimes have a curved top border, depending on “perspective”.