But in this case it would appear that formal studies were done and failed to back up the claims previously supported by self-experimentation

If you think that's the case, you didn't read the whole Wikipedia page on that, or the cite I gave to a 2001 paper that independently re-creates a portion of NLP's model of emotional physiology. I've seen more than one other peer-reviewed paper in the past that's recreated some portion of "NLP, Volume I", as in, a new experimental result that supports a portion of the NLP model.

Hell, hyperbolic discounting using the visual representation system was explained by NLP submodalities research two decades ago, for crying out loud. And the somatic marker hypothesis is at the very core of NLP. Affective asynchrony? See discussions of "incongruence" and "anchor collapsing" in NLP vI, which demonstrate and explain the existence of duality of affect.

IOW, none of the real research validation of NLP has the letters "N-L-P" on it .

You cannot verify anything by self-experimentation to nearly the same strength as by "properly validated rituals and papers". The control group is not there as impressive ritual. It is there because self-experimentation is genuinely unreliable.

Unreliable for what purpose? I would think that for any individual's purpose, self-experimentation is the ONLY standard that counts... it's of no value to me if a medicine is statistically proven to work 99% of the time, if it doesn't work for ME.