What i want to do: i want to apply blured shadows to my scene. I know how shadow maps work.
First i render the scene depths from point of light with (yes) orthogonal projection matrix to a texture. First image
Then i apply blur efect to this texture. Second image.
Then i use this texture to determine whitch parts of the rendered scene are in shadows.
In the third and fourth pictures you provide links to, it's not clear exactly what the red-and-black image is. The third image appears to be the original unblurred depth map rendered somehow with a different view, perhaps a perspective (vs. orthogonal) projection? The fourth image appears to be the blurred depth map rendered with yet a different view?
Now this is where things get complicated. The third image is renderd in the same way as the first. Same projection matrix. The fourth image is the same as the second image. Same blur effect same projection matrix. The only difference between first/third, second/fourth image is the size of texture to whitch i render. First and second textures have same size as screen and third and fourth have size 1024 x 1024.