Only R1 through R8 are used in the CRI Ra score. R9 through R15 were added later due to recognition that fluorescent and other lighting sources with spiky spectra were not easily compared by Ra to sources with more continuous spectra.
Hence also why you can have R9 (saturated red) scores that are effectively 0, yet score well into the 70’s for Ra.
Specifically regarding R12 being low, scroll down to TCS12 here to see a reflectance spectrum for the test swatch:
In short, blue-pumped LED’s have too narrow of a spectral peak to score get a great score here, as the test surface has significant reflectance extending into the violet range. To get a really good R12 score from an LED generally requires a violet-pumped LED like Nichia Optisolis. See here for example: Maukka’s 5000K Optisolis Measurements (source)
Fortunately, the typical LED scores in the R12 = 50 to 70 range keep the effect relatively subtle. Consider similarly how an R9 score above 50 is decent and above 70 is pretty good. Yet even before I understood the limitations of most LED’s at the blue end of the spectrum, I did notice this doing comparative photography tests of some of my first high CRI LED bulbs against some fluorescents. Although these specific CFL’s had a partially correctable color cast and had terrible rendering of greens, they actually did very well in the violet range. A vase of various colored flowers I used as a test subject showed the violet colors looking more blue and a bit dull.
Since I didn’t understand this at the time, I was confused that the high CRI bulb seemed to perform worse for shades of purple, but the improvements in the other colors compared to the CFL was far more significant.
As to the height of the spike, because most real surfaces reflect a fairly broad spectrum, we seem to not notice this. However, if you illuminated a surface with fairly narrow reflectance spectra centered on that same spike, the blue should appear exaggerated compared to when illuminated by sunlight.
As far as I know, human eye sensitivity to 700+ nm is so low that it is almost irrelevant. It should be possible to create a test scenario with extremely low reflectance until nearly 700 nm where a difference would be seen compared to sunlight, for example, but I think uncommon.