Indeed. The images here are mostly useless for CRI purposes. Digital cameras sense three specific frequencies, and computer monitors emit three specific frequencies. They can’t record or display the full spectrum, nor can they show a meaningful difference between three narrow-band lights and a full-spectrum light. Common image files only store three color channels, so there’s no place to put the extra color information even if it was available.
So this is pretty much all about how things look in person.
In person though, I’d take a 70CRI XP-L HI 3D (~4800K and slightly pink) over a 100CRI emitter at <4000K or >5500K. Because perfect CRI can’t fix a skewed color temperature.
OTOH, decent color temperature can’t make super-low CRI look good. Like, a narrow-band green LED may have the same peak frequency as a neutral white emitter, but it’ll make everything look terrible. Fortunately, nobody proposed that. It’s kind of irrelevant because zero or negative-CRI light sources aren’t really being considered.
As long as the CRI is high enough to classify a light source as “white”, like 70+, further CRI improvement is a relatively small factor in the result.