Colludium is right.
In WebGl rendering version WebGL can render the same textures numerous times once. Where as rendering of different textures requires more draw calls.
In C2; C2 treates one Sprite object and all the images as a single texture. So while you see the division of frames it is in fact better optimized for WebGL rendering. Where as different sprites are always rendering different image textures.
minimize the number of sprites as much as possible. In fact it's best to to creat the game with master Sprite's that are only pinned to gameplay objects. Used image sprites have collisions off, where as gameplay objects never render.
So it's always better to minimize sprite used. however C2 still needs to groups sprites by transfer size which is most of the time 2048.