I've been buying predominantly high CRI LED lamps, if i have to buy lamps at all.

One thing i've noticed is that there are two ways to get there. The first one uses some broad spectrum phosphor (1st attachment) used by a Ledvance 97 cri (disgregard the 94cri in the attachment file name!) 9,5w lamp, model number AC45033. This lamp has 85lm/w. To the eye, this one looks more neutral white but also not exactly 2700k but a higher color temperature/cooler light.

The second one, also 97 cri, uses a peaky red phosphor, likely BaTiF:Mn4+. This one is used to boost efficiency, and this lamp reaches 97lm/w. This appears the currently preferred way to make high CRI leds (as well as to get *some* red content into those 210lm/w ultra efficient LED lamps which have a piss poor (literally lol) color)

Which phosphors could have been used in the first one to reach that almost shelf-like curve between 600 and 650lm? I have not seen such a curve in any other LED lamp.

What is really funny, when you start to demand a LED lamp with 95+ CRI and a really good R9 index, lamp efficiency immediately rolls back to below 100lpw level, which was achieved some 10-15 years ago by Elite CDM lamps, having the same good CRI and R9

MH lamps in domestic use have never really become succesful due to their warm up time and incompatibility with existing fixtures. CFL won that race, despite the lower CRI for most of those.
 
But the nail in the coffin for them, is the inevitable presence of mercury and the overall expense of accurately producing metal and glass pieces of art that can withstand staggeringly high temperatures for tens of thousands of hours.

HPS lamps without mercury exist, but i'm not aware of MH replacements without mercury. But that's not what this thread's about.

You forget the main market for high-end CMH that is retail lighting. In practice, if a retail owner replaces his MH track lights with LEDs and demands no regression in light quality, there is not a much gain he will achieve. Efficiency is roughly the same, may be some gain in lamp lifetime, but late CMH weren't bad either and long life for LEDs is achievable only with well thought out thermal mode, which is not always the case)

BTW, BaTiF6:Mn4+ seems a new matrix for Mn4+ activator, like a good old KSF phosphor?

I wonder how efficient a lamp would be that uses more LEDs (like Phillips Ultra Efficient lamps do) while having a high CRI phosphor mix. I'm not sure how much of the high efficiency is due to the phosphor mix and how much is due to the LEDs being less overdriven.

I bet the problem is not with LED or phosphor efficiency, instead it is rather fundamental. Lumens per watt is weighted by eye sensitivity curve, so if you pump a lot of energy into a deep red region, this does not achieve a significant visual effect and this energy is 'wasted'. Also, some more energy is lost on the Stokes shift converting blue photons to deep red ones with significantly lower energy.

But, deep red light is present in both incandescent and sunlight and deem necessary to get high CRI and R9 scores. Cutting off deep red emission at some wavelength helps get back lamps formal lumens/watt efficiency, but the light gradually becomes uglier.
I just did not expect this effect to be that significant and that demanding full spectrum at red immediately brings LED efficiency back to the level already achieved by MH lamps 15 years ago!

Brap, this is hard to answer without knowing how hard is 'normal' driving for a certain filament, and how hard is 'economical' driving, and also what for that type of led filament the curve of light output VS input power is.
But i have a Philips 927 lamp with a CRI of 90. That one has 138lm/w.
It is a glass bulb with 4 filaments inside, running at 3,4w.

Compared to the others, it has a lot more of the red phosphor, with also relative to the main (YAG?) phosphor a fairly large amount of pump diode blue output. Perhaps it relies on the blue output to make the number go up, while relying on the red output to pull down the color temperature.

Subjectively, the Philips looks more neutral, less yellow, than the Ledvance with the extra phosphor, when lighting a purely white wall that looks white under daylight.

I'll browse the DIY store's offerings one of these days to see if they offer high CRI lamps too. I bought all of these at web shops. The Osram one is already a couple years old (i think i actually bought it branded 'osram' but received one branded 'ledvance' but i could be mistaken), the others are brand new.

In general it seems like the market has just accepted a cri of 80 as normal and acceptable, which is a bit sad. I can't say i'm terribly bothered by it, but ~ 100lm/w is already super efficient and for home use you're not gonna save a significant amount of money upping it to 150-200lm/w at the cost of proper (mostly) red reproduction.

With regards to the inevitable lower efficiency of converting short blue wavelengths to red - there are labs working on integrating actual direct red radiating LED dies into the standard blue + phosphor LED chips. This will be interesting to see, also in the sense that usually red LEDs have an incredibly narrow bandwidth and how much that detriments the overall light quality.