I have played with this.
I knew going in that I couldn't solve the optical invariability problem but wanted to play with some really high power diodes.
There is a ratio of: power, spot size, collimation, physical size. Not to mention entendue.
The size and collimation of your projected spot is directly related to to your emission (point) source. 1mm led die can't be collimated and small at the same time. Pick one, just the way the physics work.
That is the reason that they use ~25 um for fiber remote heads and take the loss. If you could use 500um AND get a small beam with more power they would.
chad
When the going gets weird, the weird turn pro.