On directivity, I don't think we should take for granted that wide dispersion is always better - directivity is perhaps the area individual preferences most come into play (also contingent upon your room).
From my recollection seems
most people prefer a wide directivity design (I certainly do), but the research also makes it pretty clear that some people prefer narrow directivity, especially when it comes to music production. From Toole's book, section 7.5, this referred to a test performed with professional engineers:
In the control room, though, only about half of them felt that they could produce recordings with wide-dispersion loudspeakers. So, most of these professionals liked room reflections for recreational listening and about half of them thought that they could mix with wide-dispersion loudspeakers and the lateral reflections that resulted from them. Clearly there are individual differences.
This relates to how we should deal with sidewall reflections as well. There is some room for preference and room-matching here.
Back to the Buchardts, as someone else noted, its big crossover dip comes from the floor reflection. From my understanding, this reflection has perhaps the smallest influence on our perception of sound quality. Posting the reflections curves again for reference:
From section of 7.4.7 of the ubiquitous book:
Very early in my explorations of loudspeaker/ room interactions I took note of the measurable effect of the floor reflection in steady-state room curves. It seemed like a problem that needed attention, so I devoted some time to modifying loudspeakers to minimize it, and followed through with subjective evaluations. I cannot claim to have been exhaustive, but I soon became frustrated when the curves looked better but the sound seemed not to have changed very much.
He elaborates further in that section, noting that Linkwitz found the floor reflection is "not necessarily audible" on music, and one study suggested removing the reflections altogether has a detrimental effect. It seems we've adapted to largely ignore floor reflections because it's always present in our lives.
This isn't to say that the floor reflections shouldn't be optimized. They presumably impact the early reflections spins for a reason, and it seems it's a topic that merits further research. But it does seem like if the Buchardt's crossover dip is emerging from floor reflections, it is likely not as much of a problem as it might seem.
On the other hand, ceiling reflections
are problematic. While sidewall reflections tend to contribute to spaciousness, ceiling reflections tend to be perceived as timbre changes (section 7.6.5.1) due to spectral differences, which is not good.
If Buchardt decided to sacrifice floor reflections for better ceiling reflections - and indeed, its ceiling curve looks almost as good as its sidewall curve - this seems a reasonable choice.
Edit: That said, I have heard of some research suggesting the contrary. It seems floor reflections merit some more research. It seems to me like wide horizontal directivity, narrow vertical directivity is good on non-coincident designs.
