I noticed that the datasheet shows a cutout through the LED. It shows that the die with phosphor is about 0.06 mm thick mounted on a pole.
This allows me to calculate the side area which may be a part of the LES.
That extra area is about 0.18mm².
If sides had the same luminance as the front this would account for ~27% luminosity gap against theoretical calculations. In the actual measurements the gap turned out to be ~33%.
Now…I would be surprised if the sides were the same luminance as the front. But that would explain the discrepancy.