To put it in a few sentences:
The engine isn't re-calculating the size of a bitmap at all.
It simply places it on the surface, and then(!)
it translates the places of the pixels to the pixels of the screen, depending the position and arc of the camera...
The reason for unsharp bitmaps is the - intentional - mipmapping.
This means the pic is precalculated in different sizes, where the engine chooses the appropriate size depending the distance and angle of the triangle in relation to the screen.