Conversion between overall output (lumens) and intensity (lux, candella, footcandles) is not easy.
If I recall correctly, a lumen is 1 cd (or lux) per steridian within a sphere and there are 4 pi steridians per sphere.
I once devised a crude
spreadsheet for converting mcd readings with a radiation pattern to lumens. It needs some work to simplify the calculations. One cell to calculate that all will be messy; perhaps I could get ambitious and get it down to a single row of calculations since many of those cells are static values I determined graphically, some are now vestigial, and some are both. Feel free to use or even improve upon it; I only request that you notify me of any improvements and send me the improved file.
As far as using that spreadsheet goes - enter max and minimum mcd for the LED (or put the nominal mcd twice to calculate and absolute rather than average) then look at the radiation chart and enter the relative intensity for each 5-degree sample point. The example in the spreadsheet is a Nichia
NSPW515BS.