Related to one of Greg’s “watch-it” comments about a photometric report being relative to lamp lumens - this is less common now, with LED data mostly being reported as “absolute” photometry not relative to lamp lumens. This is largely because so many LED products are really lots of tiny LEDs rather than one main traditional lamp.
If your graphical plot shows total candela from the source at each angle (doesn’t need any scaling by lumens), then enter the values into your invented IES file according to the plot, and enter a value of -1 for lumens. That is the marker to indicate it is an absolute test (not relative to lumens).
I don’t think the watts value would impact your Radiance definition. It think that’s only used in software like Dialux or AGI when they have lighting power density functions (maybe in things like Open Studio too?). I’d just put in a value that’s your best guess of the light source input wattage. If it’s an actual candle then I guess it’s not connected to power. I’m not sure but maybe the file format won’t complain if the value is a 0.