a neopixel or alike takes 3.5ms to refresh a 120 pixel strip (on my boards at least). That's 285 fps while the eye can deal with 30-60.
The APA102 or alike can have a faster refresh rate but overall they are mostly used because you can refresh the strip at the rate you want, with less time constraint.
Driving a neopixel strip leads to more intricate programming if you want it to have no impact on the latency of your board, and I've see one board so far using it "by default", simply because the dev couldn't deal with both sound and refreshing the neopixel, which is super CPU intensive if you handle it "the simple way". As pointed by adafruit, neopixels can feature a maximum of 400 Hz fps which is not suitable for PoV systems (which a saber isn't, even if the Profezzor'n demoed it's clearly possible) and are usually used on large number of pixels (I have friends in performance arts refreshing thousands of leds at once at high FPS).
In the same idea, neopixels would be very limited to display video for instance, as soon as the res increases, your refresh rate would drop on 1/X^2 which is quickly catastrophic (same for PoV) reason for increasing the pixel refresh rate like the dotstar proposes.
I've experimented with the first generation of dotstar strips, they "work" without a doubt. But I'm just happy with my zero-CPU neopixel driver, I like the simplicity of it, then it uses only one data wire which is also helpful for the connector system. Dotstars also have more burn-out along the strip (1 additional data wire on the same width leads to less copper density, or maybe it's just the manufacturing). This last point is less of a concern on a 90cm strip though.