R9 is also important in the higher CRI ranges. CRI is based on pastels - it's an artifact of measuring traditional florescent bob performance - and R9 captures the vivid reds that are important to truly accurate color rendition.
Just from the scale I think it's clear that CRI is not that good a measure. As you say, CRI uses rendition of a dozen or so colors as a measure. The scale is heavily compressed - orange sodium vapor streetlights which are almost monochromatic "achieve" a CRI of 30, which is nowhere near "colors are 30 % good". Similarly, mercury vapor lamps have a CRI of around 50, and yet are complete garbage. All cheap LED fixtures are CRI 80 or 85 which is also garbage. 95 and 98 sounds very close, but aren't.
SSI - spectrum similarity index - is probably a better measure, because it doesn't look at a few colors, but at the entire spectrum, and so heavily punishes peaky and discontinuous spectra as produced by many LEDs. High SSI values also result in much better mix-and-match behavior of light sources. Even CRI 95 LEDs often don't mix well, usually on the pink-barf-scale (tint).
