The casualty radius is very different from the lethality radius that is described in the article - in military terms, a ‘casualty’ is any injury sufficient to have a soldier pulled from the battlefield, so a unit could suffer 50% casualties, and 0% mortality in a combat. If I roll out of my bunk and hit my head on the floor and concuss myself during a surprise attack, I’m a ‘casualty’. Maybe not getting a purple heart for that one tho…
Thus, the 15m casualty radius is the radius within which you’d be expected to suffer one or more injuries sufficient to potentially remove you from combat - very different from a lethal radius, especially given that our Survivor frequently is forced to fight through wounds which would be considered ‘casualties’ in warfare, as they don’t really have a rear line hospital to be carted off too (and they heal far faster than a real soldier, so the concept of a ‘casualty’ is very different in CDDA). Obviously you can still certainly be killed within that radius, with the equivalent of a ‘critical hit’ to a vulnerable spot such as an eye or vital organ, but generally speaking you’d just be suffering flesh wounds of a significantly lesser variety to those of small arms at the same range.
In terms of the physics, both the thinning of the particle cloud and the reduction in velocity are going to be relevant, though It’s much harder to calculate the latter without knowing specifics. The fragmented pieces will vary in size considerably and will thus vary in how fast they are moving initially and how quickly they slow down to terminal velocity. You’re more likely to be hit by smaller fragments, as they’ll be more numerous, but the injuries they cause will be lesser, and they’ll be more easily stopped by armor. The large fragments will be more dangerous AND will go further, but necessarily much fewer in number, so your odds of being hit by one outside of the first several meters is very low.
Under no circumstances should we expect any fragment from a standard grenade to have the velocity or momentum of, say, a standard .38 pistol round. The explosion is unconstrained and thus unable to efficiently accelerate the fragments. By the time the fragments are just a couple inches out from the initial center of the blast, the bulk of the explosive force will simply go around the fragments, quickly leaving them behind and failing to accelerate them much further. Most of the blast force is simply wasted, unless the target is close enough to suffer the direct effects of the shockwave, such as deafening and dazing - if they are close enough to suffer concussion and internal shock injuries from the blast wave itself, they’re probably inside that 95% lethality radius.
The reason that lethality drops off so incredibly fast for a grenade is that it is dropping off on several axes at once. The fragment cloud density is dropping with the square of the radius, the fragments are losing momentum at a drag coefficient far higher than that of a more massive, aerodynamic bullet, and all of them are suffering drop, which means that past a relatively short distance the velocity loss of fragments that can hit a target is even greater than the straight line distance indicates, because the only fragments reaching out past that first 15 meters or so must follow longer ballistic trajectories. This last matter may sound insignificant, but it is a large part of why many traps and bombs are set to air-burst, like bouncing betties. It allows a far larger proportion of the shockwaves and fragment cloud to engage targets at a ‘significant velocity’. Whereas a normal grenade going off on the ground is completely wasting at least 80% of its fragmentation volume, 40% being shot straight into the ground under it, ~40% being fired up into essentially harmless ballistic trajectories (as long as you aren’t standing on it when it goes off), and only a relatively thin ring around the edges moving on vectors that will directly threaten the intended victims. Air-bursting allows most of that lower 40% to come into play if a grenade goes off, between, say, 5-10 meters off the ground.
I think the big rule of thumb you should keep in mind is that in any kind of blast fragmentation, you should expect the fragments to AT BEST behave like a similar gauge of shot from a shotgun - in reality their scatter is far wider, and for any given fragment mass, their drag will be greater than the equivalent mass of shot. As far as initial velocity goes, it obviously depends on how much explosive was used, but generally speaking packing a lot more explosive behind the fragments has a very bad return on investment past a certain point (because most of the blast force is necessarily wasted due to the lack of a pressure containing mechanism such as the shotgun’s barrel), so realistically they just won’t be going that fast compared to a bullet from a gun or a bb from a shotgun.
Correction: I believe I was correct in saying that you see a bad return on investment for piling on more explosive past a certain point (in particular the point at which the blast factures the container) - however, the initial velocities of the fragments CAN be very high, because true explosives like TNT are much more energetic than standard gunpowder. So yes, your fragments might be going faster that bullets depending on details - but no, they won’t be doing so for long, as their drag coefficients are far higher than bullets.
Here’s a fascinating source on the subject, if a tad hard to read due to bad army copying machines. 
