Prof - I really appreciate you sharing your 2.7/3.7 selection journey. Your lived experience with those models is extensive; mine is not. Your observations parallel mine from a single audition.
Speakers are very complex organisms. Each and every observation results from many interacting variables. The 2.7 face some challenges that would not have been on Jim's plate: namely the crosspoints are different than the driver design(s). The bass-mid crosspoint of the 2 is an octave higher than the 3. Among the most sonically important differences is the very large capacitance needed for the midrange to roll in higher than it wants to. 2.7's 416uF vs 3.7's 226uF-all polypropylene. 400uF of the 2.7 is a single electrolytic cap. Jim never exceeded 100uF for good reason.
As we all know, descriptive sonic vocabulary comes up short. I'll resort to a visual analogy. Among the aspects of light description are 'balance' and 'rendition'. Balance, described in degrees Kelvin, is much like sonic tonal balance, how warm or cool. Color rendition describes the spectral content benchmarked against sunlight, which by definition illuminates all wavelengths equally. A low Color Rendition Index notes incomplete wavelength content. Frequencies that aren't illuminated can't be seen.
I liken a large electrolytic capacitor to low CRI. I'll call it SonicRI. In the 2.7, some of the sonic content, mostly within our envelope of 'detail', is obscured, swamped by parasitics in the cap. Your descriptions of imaging shortcomings fall in this arena. I found the 2.7 to lack the fine-grain resolution of the 3.7, especially in the time domain. I speculate (I haven't measured a 3.7 or 2.7) that zooming in on the 2.7 onset transient of the impulse / step response would show a slower and less ideally shaped rise. Also, the integrity at the top of the spike would be less organized, more ragged - less precise would be my summary. This misbehavior would be masked by the better tweeter behavior, but it can be heard as a defocused sheen across the midband.
To your question. Some users here have replaced that 400uF cap with 4x100uF and report cleaner sonics. They might chime in here. My further refinement uses 8x50uF plus a more sophisticated bypass cap, all in a unified concentric field. Field trials are underway in the SCS4 and CS7.2. There will be more actual outcomes to report shortly.
Feel free to PM to discuss further.
