Yeah i get it and when a grenade explodes in the corner of a room, a smart algorithm should create a decal mesh that covers e.g a piece of the floor, the front and the right wall. isn't it.

Things will get a lot harder when the decal must be mapped on different objects e.g the terrain and a rock. Apperantly you should create an algo that determins the most important mesh(es) to apply the decal on.
But i dont think there are any multi-object-decal games/engines around yet... because it must be very computational intensive.
Having nice 'parallax-mapped' decals like HL2 does, seems sufficient to me.