Hi Jimmy,
All parts will have variations and are typically given a tolerance range. With something like a parabolic or other curved surface, I admit to not knowing how the "professionals" would go about measuring a part and confirming whether or not it is within tolerance. I asume that there are 3D tracers which can follow a surface and report the path to a computer. The companywho makes the reflectors for me have been real good about holding the tolerances I have specified but I have no way of knowing how close they come to the curve and location of focal point or how this may vary from part to part. I would guess that the variations in film thickness and surface orange peel in the reflective coating would be the greatest contributor to visual differences in beams using the same LED.
I have been told by a friend who would know, that the distribution pattern of light from one Luxeon to the next can vary significantly and to an extent similar as what we have seen in flux, Vf and tint. Die centering, elevation and distribution of light from the dye all likely contribute to differences in beam.
A computer modeled optic or reflector that is based on a specific distribution pattern and not very forgiving in terms of misallignment likely has trouble in real world applications if the emitters it is paired with in fact vary to significant levels in deviation from the presumed distribution patterns. I have heard rumblings of this being the case with the LED's. :shrug:
With the reflectors I have designed, I have intentionally gone for forgiving reflective surfaces which can accept variations in allignment and part dimensions and still provide an artifact free beam.