I was wondering the same thing and finally found the answer:
Visually, the two modes have nearly identical behavior, which is why it's confusing. As you noticed in both modes, the window you're dragging will take up the full, half, or quarter of the screen.
But when you use the modifier key (Ctrl by default) to switch to snapping mode, the window you snapped will be considered "part of the screen real estate." That means other windows you maximize will try to avoid the space occupied by the snapped window. If you tile rather than snap (i.e. the default if you don't use ctrl), maximized windows will simply cover it.
So the difference between the two modes is not really with the behavior of the current window you're snapping, but with the behavior of other windows that you maximize later on.
You can hold a modifier key (defaults to Control) to toggle into
“snap” mode – this is unique in that a snapped window will be
considered part of the screen real estate, and maximized windows will
avoid snapped windows. Tiled windows (unmodified) are treated just as
before – they are ‘stuck’ to the screen edge, but don’t receive any
Tested and confirmed on 19.1 Tessa