Realtime resolution vs. baked resolution

Can someone explain to me how the realtime resolution works?

The result in setting both to 1 makes a huge difference and I want to understand where the difference comes from.

This is how I understand the baked res:

object surface * baked resolution * Scale in Lightmap

e.g. a plane with 10 units length * res of 1 * scale of 1 = 10 * 10 pixels lightmap

e.g. a cube with its 6 faces * res of 3 * scale of 1 = 18 * 12 pixels lightmap

The resolution is way too low to get the same result as with the realtime resolution. But even with the same value 1 for realtime res, realtime shadows look way better. How is that possible? Is it internally boosting the scale and lightmap size on its own?