In my experience, it is much better to hybridize Ha with R, rather than a standalone channel in RGB combine, or with a luminance image. This produces much better color in the final result, the others tend to produce the sickly salmon color you often see in Ha combined images. I always normalize the color space with RGBWorkingSpace prior to splitting into R, G, B and combining the Ha with the R. I never use L*a*b color space in this process.
It varies from image to image, but play with the percentages of Ha and R in the hybrid, then recombine to form the new HaRGB and assess. Sometimes more Ha, sometimes less....