4895 is a relatively faster burning powder by rifle standards and makes more pressure than most bottlenecked cases can handle, if loaded to a full case, but does really well in terms of consistent ignition and velocity, in most cases, if loaded to the level that gives safe pressures. It is probably at its best in cases like the .223 and a bit fast for bigger cases and a bit slow for smaller cases (speaking in overly generalized terms of course - the real issue isn't case size, it's the expansion ratio).
There's nothing magic about it - most powders have a range of pressures in which they'll ignite consistently. H4895 just happens to do so in a range that lets it be really useful for reduced loadings. I have used IMR4895 in a similar fashion and while Hodgdon recommends against it, I suspect that's more due to a lack of data, than any actual shortcoming with the powder itself. Also, my 'reduced' loads aren't at the ~60% level people refer to for H4895. My main use case for 4895 is only 'reduced' in the sense that they're less than maximum loads in an effort to get, say, 90+% of the performance of any given caliber with 80% of the recoil, which is a trick you can do with a lot of calibers in order to move yourself around on the performance-to-recoil scale.
When I was a kid my grandpa let me shoot a ton of his old surplus .30-06 ball ammo and when he 'sold' me that rifle he also loaded me a bunch of ammo with, IIRC, either 49 or 49.5 grains of 4895 (don't remember H or IMR) and a Sierra 150 grain Gameking. I killed my first couple of deer with those slightly reduced loads. They didn't seem to notice the reduction, and thus I got a lesson in recoil that took another couple of decades to begin to properly sink in.
