I got a promising candidate as a substitue for the GUVA-S12D. I have a 405nm UV LED which tracks a GUVA and VELM sensor in daylight well. It has a further advantage of outputting a voltage almost in range of the 2.56V internal reference. I find a 1 M resistor in parallel with the LED is enough to pull down the voltage within the upper bound of 2.56V.

This will enable me to get the parts count down siginificantly. I can drop the op-amp and the associated passives. My LED is $0.18 in quantity.

I noticed it is more sensitive to my 385nm LED flashlight than the GUVA, so I'm using real sunlight to gather readings. I'll need to wait for the sun to get higher in the sky (or a sunny day in the mountains) to get up into the 9-11 UVI range. 11 will be enough to get to the "extreme" range. I plan on simplifying the readout into the four cateories of "low", "moderate", "high", and "extreme".