Just to better illustrate what happens with <60FPS and vsync:
http://james-ross.co.uk/temp/orts_124.png
This is Open Rails running the same scene (so 40-45FPS) full screen (with fast full screen disabled) with vsync enabled. In the "Flip Queue" you can clearly see that the time between a frame being finished (and thus the present making it in to the queue) and it being presented (at the vsync) varies wildly. The presents from the game (see the yellow boxes in the "Context CPU Queue") are evenly spaced out, however, so you get a mismatch between the game's idea of time between frames and reality (this is what Wayne is talking about). There isn't much you can do about this, apart from disabling vsync (which gives you tearing instead), or G-SYNC.