I used a calibration lamp inside the sphere in a E27 holder. After that I put the Emisar D4 in there at 100% 7135 for flashlight calibration (subtracting the effect of the D4 itself using an auxiliary light). After that I moved the D4 outside the sphere and measured the lux through a diffuser on the input port (40/100mm adjustable). I’ve noticed that a diffuser is the best single thing to improve the integration.
Now I had a calibration light for measuring other lights. For them, I then use the aux light again to subtract their effect on the readings. But you know all of this of course. With the diffuser in place, the small (<30mm) lights don’t really affect the readings, but for bigger lights such as the X7R the effect is clear (bigger readings when the reflector bounces light back to the sphere).
Of course the actual process involves tons more testing and verifying with several lights, but that’s it in a nutshell.
Sphere pictures are here: Imgur: The magic of the Internet