The calculator is based on using white LEDs only, and they tend to produce about the same lumens for every version of white. If you add in a lot of red/blue, etc, LEDs the calculator won't give good results. This is an empirical calculator, not a theoretical one. It is based on using a variety of data and finding an equation which works well enough with all of the data.
This is an inherent problem with LED calculations.
The more light type variables you add, the messier things start to get. There's much less consistency since you can DIY so many options.
Still, you can get reasonably close with the Whites, then go from there when you add say a set ratio of red and blue mixed with a % of the white.
Say 10% blue, 50% red, 40% white.
Seems that most of the blue types will be close to one another(I would assume), and likewise the same for the reds.
You could add an estimation of those 3 variables without too much extra effort. Then real testing to see how far off the model is/is not. You need some data: averages of the reds and blues for that, then some averages on real fixtures to see how well the model works. If the averages are a bit off, you can then go back and add a correction factor to make up for that in the model.
A few step wise adjustments, I think it can work out fairly well to account for the other colors. Since many folks ALREADY have white/red/blue mixes in their fixtures, you just need some of their PAR data basically.