The graph states lumen per watt, not percentage efficiency.

Indeed. Theoretical maximum is over 200 lm/W: https://en.m.wikipedia.org/wiki/Luminous_efficacy

Oh, that is nicely misleading. The combination of being totally awed by the close proximity to perfection implied by the misreading and being used all of SI fitting together like Legos completely got me (see downvoted comment below, I guess the yaysayers could be right once more - and even now I still think that a photon stream worth one Watt of power, constrained to the visual spectrum could be a very nice alternative unit for the value usually given in lumens).

And yet you won't find any shortage of people who will call you a terrible, terrible naysaying pessimist if you dare to voice the prediction that an efficiency jump like from incandescents to CFL or from CFL to LED won't ever happen on top of state of the art LED.