Found this cool article, looks like they have now achieved 50 lumens per watt and they think they could even reach 100 lumens per watt. The EISA 2007 bill requires that general service lamps must meet a minimum of 45 lumens per watt by 2020. Check it out:

http://www.hybridlightbulb.com/index.html

There are two hypothetical ways to improve efficacy of incandescent:
Increasing temperature - then there is a limit of (if i remember well) 70lm/W with infinite temperature, i guess all know, not as practical solution for further increase, as the limit for usable lifetime (~1000..2000hours) was already reached.
Second option is to reflect unwanted wavelength back and use their energy to aid with filament heating (used e.g. on Halostar IRC from Osram). But this mean, then a kind of filtering layer and accurate bulb shapes should be used. The second is easier to get today, but the first would bring with it unwanted behavior: Affecting lamp's light output by parasitic reflection (or absorbtion) of some part of visible light, so reduce CRI. Be sure, there would be same compromises between efficacy and CRI as we see today on fluorescents. And i'm sure, necessary materials for such filtering mirror would be anything, but nontoxic, so we endup with same problems as we see with fluoros at their EOL: Hazardous waste. And no manufacturer of "high efficacy incandescent" would run them cooler (on nominal voltage) then required for ~1000hours life, simply because the first point about incandescent efficacy is still valid.
So i think direct conversion based electron energy level change yielding a photon (so all positive column discharges and LEDs) with phosphors would still win the battle for best CRI*Efficacy combination.

Found this cool article, looks like they have now achieved 50 lumens per watt and they think they could even reach 100 lumens per watt. The EISA 2007 bill requires that general service lamps must meet a minimum of 45 lumens per watt by 2020. Check it out:

http://www.hybridlightbulb.com/index.html

I'm wondering, what's the CRI of halogen infrared lamps such as in the Philips Halogena Energy Savers and GE HIR PAR-38 lamps? You said they have a lower CRI than non IR lamps, yet others still say they have a 100 CRI, so I am just wondering. Thanks.

I'm wondering, what's the CRI of halogen infrared lamps such as in the Philips Halogena Energy Savers and GE HIR PAR-38 lamps? You said they have a lower CRI than non IR lamps, yet others still say they have a 100 CRI, so I am just wondering. Thanks.

The CRI depend on the "cut-off" wavelength of the reflecting mirror. Today, on comercially available lamps, it is at really long wavelengths, so the CRI is indeed practically 100, but the efficacy gain is only ~30%.
The question is, how far to go:
Or move the border to about red part, so the efficacy might rise significantly, but the light would be defficient on red, so the CRI would not be 100 anymore. I don't think it would be possible to make efficient filter in such a way, it would gradually pass more and reflect less (when shifting to blue) and absorb nothing in the way the result of reduced red would be increased CCT and stil CRI >90. Most likely such creation would have or strong greenish tint - red reflected back to filament and blue not emitted strongly enough (remember "natural" incandescent spectrum) or again lower efficacy (the "filterring reflection" would have to extend to green, but there would be significant absorbtion as well)
Or the border would not be moved that far, so still significant amount of IR would not be reflected back.

Note, then ideal "filterring mirror" (reflect all longer then the border and let pass shorter then something, while absorb nothing) would never exist, all would have quite wide band of absorbtion (reflect, then absorb, then pass), so or all the IR would stay in the efficient reflection area, but the absorbtion would extend to visible, or it would be dimensioned so the all visible is in "pass" area, but then significant amount of IR from the filament would be in "absorbtion", so not reflected back.