I have found in my non-scientific, small batch testing, that crimping my 5.56/.223 rounds to be used in my ARs reduces SD enough that it should/could be conducive to better groups.
With that said, if the bullet does not have a cannelure I don't crimp as I think deforming the bullet is worse than not crimping. But if the bullet has a cannelure, I will generally crimp that round.
I usually don't use bullets without a cannelure in my ARs. Each to his own.
I think if you pay close attention to neck tension, crimping is not necessarily needed. But I, and I'd bet most guys that reload 5.56/.223 range brass don't have a clue what the neck tension is throughout the group being loaded. I think unless you anneal, even once fired cases will have very different amounts of neck tension. I believe it is this widely differing tension that will increase SD and probably some malfunctions.
I prefer once fired LC brass or commercial once fired, (but you never know if commercial brass is once fired). But I load whatever is in the batch, of course separated by head stamp.
I'm sure you have experienced different brass resizing harder than other. That will effect brass prep through all steps including the end neck tension.
Sorry for the long post but I think many would benefit rethinking crimping in general. YMMV