This morning I rechecked the reading vs a Quantum PAR meter. It read 21 with the PAR meter reading 24-25. Close enough for my purpose. I will recheck it again in a few days to verify that the reading is stable.
I also made another try at selecting filters the give a more even spectral response. It came out worse than what I used. The difficulties are finding a filter to knock off the near infrared and near UV, without distorting the rest of the spectrum too much. Then finding another combination to even out the response, without getting too many peaks, and still another combination to drop the sensitivity sufficiently without really exaggerating the peaks. I keep thinking it can be done better, but I'm out of ideas now. My "obsession" has about run its course now.