It is true that main perceived benefits for alternative layouts are a long-term affordance rather than a peak speed, but in my experience the threshold is pretty high, if not non-existent (the professional transcription typically uses a dedicated keyboard layout with lots of common phrases anyway, so the general-purpose layouts are not relevant there).
I'm not a proficient English writer but I had been in a similar debate for Korean keyboard layouts, with a major split between consonant-vowel models (두벌식 "bipartite type") and initial-medial-final models (세벌식 "tripartite type"). For your information, a single Hangul syllable consists of one initial consonant, one medial vowel and an optional final consonant, with clusters possible for each part. Tripartite models can be (at least theoretically) more accurately optimized as frequency distributions for initial and final consonants would be pretty different, at the expense of slightly more keys to fit in. This theoretical consideration, combined with the fact that distinct keyboard layouts were directly tied do distinct typewriters in 70s and standardizing on the existing layout would effectively encourage a single typewriter, made the Korean government to standardize on entirely new layouts [1] with a bipartite basis. Nowadays tripartite layouts are supported in all major OSes but users are hard to find.
I had been a reasonably fast typist both in English and in Korean. I can continuously type Korean at sustained 600--700 strokes per minute [2], both in the standard bipartite layout and in a well-known tripartite layout (though I do feel pain after typing too long in the former). I can also type English at sustained 90--100 wpm in QWERTY. For both cases these peak numbers were highly meaningless even in the most favorable condition; a brain is not that quick. I had even done various transcription works from time to time and the main limiting factor was a recording quality and not the typing speed. (My work is by no way professional, but I have heard that one task was also sent to a professional transcriber and he/she had gave up due to the recording quality.) That's why I finally gave up learning DSK or elks.
[1] The standardized layouts were different for typewriters and computer keyboards; two are similar enough to be considered as a single layout but the former is typically termed quadripartite rather than bipartite.
[2] "Strokes" because consonant and vowel clusters often take multiple keystrokes. Roughly comparable to 100--150 wpm in English. Of course, all these numbers are not directly comparable to those in the typewriter era.
I'm not a proficient English writer but I had been in a similar debate for Korean keyboard layouts, with a major split between consonant-vowel models (두벌식 "bipartite type") and initial-medial-final models (세벌식 "tripartite type"). For your information, a single Hangul syllable consists of one initial consonant, one medial vowel and an optional final consonant, with clusters possible for each part. Tripartite models can be (at least theoretically) more accurately optimized as frequency distributions for initial and final consonants would be pretty different, at the expense of slightly more keys to fit in. This theoretical consideration, combined with the fact that distinct keyboard layouts were directly tied do distinct typewriters in 70s and standardizing on the existing layout would effectively encourage a single typewriter, made the Korean government to standardize on entirely new layouts [1] with a bipartite basis. Nowadays tripartite layouts are supported in all major OSes but users are hard to find.
I had been a reasonably fast typist both in English and in Korean. I can continuously type Korean at sustained 600--700 strokes per minute [2], both in the standard bipartite layout and in a well-known tripartite layout (though I do feel pain after typing too long in the former). I can also type English at sustained 90--100 wpm in QWERTY. For both cases these peak numbers were highly meaningless even in the most favorable condition; a brain is not that quick. I had even done various transcription works from time to time and the main limiting factor was a recording quality and not the typing speed. (My work is by no way professional, but I have heard that one task was also sent to a professional transcriber and he/she had gave up due to the recording quality.) That's why I finally gave up learning DSK or elks.
[1] The standardized layouts were different for typewriters and computer keyboards; two are similar enough to be considered as a single layout but the former is typically termed quadripartite rather than bipartite.
[2] "Strokes" because consonant and vowel clusters often take multiple keystrokes. Roughly comparable to 100--150 wpm in English. Of course, all these numbers are not directly comparable to those in the typewriter era.