The more complex B4 and P4 sequences, although theoretically superior, do not perform well in practice.Oh well. It can be hard not to take this personally, and feel, after long days in front of the blackboard/web browser, that as a researcher and even a person, I am "in practice quite poor," "less useful than initially expected" and even "highly sensitive to the presence of off-resonance and phase errors." Happily, though, I'm still "theoretically superior"!
On the other hand, I did recently propose a "bubble-collapse" theory that explained the noise electric kettles made, and after some pointless arguing about rival theories, designed and carried out an experiment that proved I was right. (Stir it and the noise goes away!) Nevertheless, it's probably good that I've moved to CS.
Beware the Phase Errors, my son!
The sigma_X that bites, the sigma_Y's that catch!
Beware the Homogeneous Broadening, and shun
The Far-Off-Resonance Bandersnatch!
p.s. If you're on a job committee for one of my collaborators, I should point out that our sequences are just optimized for one kind of error, and of course by ignoring others the practical performance will be worse. Our paper should be thought of more as introducing new techniques/frameworks for producing composite pulses than as providing ready-made sequences that can be put into experiments. But please read it yourself if you're not sure.