For road traffic, the graphics scale means that the length of road vehicles is very significant compared to their braking distance in the physics scale. Hence the capacity of a Simutrans road is much less that the capacity of a real road rescaled according to the physics/economy scale. Even worse, it is currently not possible for multiple vehicles to occupy the same tile, which effectively sets the spacing between vehicles to be at least the (physics) length of a tile.

On the other hand, in real life the braking distance of a train is significantly longer than that of a road vehicle, and must be reserved exclusively for that train (including up to and slightly beyond the next signal following the braking distance). Hence the length of a train is much less significant in determining the capacity (in number of trains) or a section of plain line. (Things are a bit different at very low speeds, or around junctions where it matters how quickly/easily a train can stop blocking an intersecting route). This means that in Simutrans the capacity of a rail route is probably similar to the capacity of a similar rail route in real life, after rescaling according to the physics/economy scale.

I'll try putting some rough numbers into this. I'll use (approximately) the Extended default scale for these numbers, since I'm a bit more familiar with that.

a) Trucks travelling at 60mph (27 m/s)

Real life: Let's take the spacing recommended in the UK of 2 seconds between consecutive vehicles. Then each vehicle needs a space of about 54m, plus perhaps an allowance for the length of the vehicle, so maybe 65m

Simutrans: In Extended each tile represents 125m (for physics/economy). So we need at least 125m per truck. In practice we need a bit more space because the vehicles won't be perfectly spaced, and (depending on physics model) will either be stopping instantly and the losing time accellerating when they get too close to the vehicle in front, or will need to allow enough additional space to stop before reaching the next tile, until the vehicle in front leaves it. The stopping distance of a truck at 60mph (in real life, which should match the distance in the physics model in Extended) is approxiately 110m, of which 40m is reaction time and 70m is braking distance. So we potentially end up with a spacing of over 200m per vehicle, which is less than a third of the real life capacity.

b) 400m train travelling at 60mph

Real life: Braking distances vary a lot, and I've found figures ranging from 400m to over a mile. So instead I'm going to assume a headway of 2.5 minutes (which seems to be not uncommon, and will take into account infrastructure limits, etc.). Then we have a spacing of about 4000m.

Simutrans: Assuming the physics is the same, then the main difference is the discrepancies in train length. Working with a scale of 10 units (10/16 of a tile) per 20 metres, we get a total length of 12.5 tiles, which is about 1600m. This means a 400m (graphics) train takes up 1200m more physical track than in real life, and hence we will need to increase the spacing between trains to about 5200m. This corresponds to a reduction in capacity compared to real life of less than 25%.

In practice, in Standard it seems to be common practice to lay signals far more densely than in reality, which allows trains to run more closely together and mitigating any loss of capacity from the mismatched scales. (Some paksets might underestimate the cost os signals, Simutrans does not simulate the impact upon reliability of having more signals, and Standard does not simulate the need for additional aspects or lower line speeds when signals are closer together).

Also, note that the above assumes that the physics/economy scale is only 125m per tile. In Standard I think it is 1km, and so the scale mismatch even more heavily penalises roads compared to rails.