There are a couple of observations that may help.
First, this is a straight razor edge, so there should be a single bevel of angle around 16.5 degrees.
After the Shapton 8k, the bevel is indeed 16.5 degrees and extends all the way to the apex and there is an edge width of between 100 and 200nm. There is no convexity.
After only 10 passes on the 0.25 micron diamond, the last 10 microns of the edge is convexed (or micro-convex) and a there is a thin foil edge remaining. The edge width is around 100nm where the foil is broken away.
After another 90 passes, the foil is mostly gone; however, there is little additional steel removed (compared to 10 passes). The edge width is around 50nm. The micro-convexing action occurs within a very few laps, and from there is edge shape is “stable”
The depth of the scratches is much less than the size of the grit. The 0.25 scratches are 10 or 20 nm deep. The “deep” scratch from the “large” particle shown in this image is no more than 100nm deep; however, it does not extend to the apex. I would speculate that because of the micro-convexity, the scratch ends a few microns from the apex – in other words the apex is above the plane of the bevel and the “large” particle does not touch this last few microns.
The percentage of “large” particles is not very high, so there are very few of these “deep” scratches. If large means 1 micron and deep means 100nm.
.
.