You're correct. The shape of the tool is the way it is so it doesn't get wedged at the bottom of the seal groove.
The part in contact with the old seal is about as wide as the seal, the part in contact with the seal groove is narrower, as it should be.
The tool as shaped by the OP would get wedged between grime and fragments of the old seal or might cut the old seal because of the too narrow point that only contacts the old seal in the middle or on either edge.
FWIW, the seal I had to pull (the PS pump one, not the one mentioned in the original blueprint) has space behind it for a good catch, so I didn’t run into any problems with the tool as I made it. Lucky, I guess :)
So here's a puzzler... I think you are backwards. Based on what I've seen , the arrows show the direction of sight, so the part would taper towards the inside of the hook. It's not clear how this would be gradual, either, i.e. tapering in two dimensions, so the drop-off would be sudden. I'm still trying to figure this out.
The original 2D drawings are in first angle projection (inferred). The drawings in your links are third angle (also inferred). This will affect the placement of the section view, but will not change how the section view looks like.
i.e. in 3rd angle projection, the section view would be on top and not at the bottom of the parent view.
Hopes this helps to clean up some of the confusion.
The alignment of the section view is off as well. It should be aligned with the parent view:
Very nice write up, and it seems like you had a lot of fun.
I play with OpenSCAD for fun, but because I use Inventor and Rhino at work, and FreeCAD at home, I don't use OpenSCAD to actually design stuff.
If I want to do generative, creative and coded design I use Racket with the Rosetta package  with Rhino or AutoCAD as my backend. I am trying to port Rosetta to FreeCAD , so it is all free. Racket with Rosetta is a lot more free or open to creating generative designs than OpenSCAD. I really get a kick out of Grasshopper  with Rhino too.
> I am trying to port Rosetta to FreeCAD ... so it is all free.
This sounds really interesting. I've been wanting to work with Rosetta but with a backend other than AutoCAD, hopefully something that could run in Debian. I'd be interested to follow any progress you make on this front if there's a code repo / link you could point to, or maybe message me (contact info is in my profile).
Rosetta already has an OpenGL backend.
My job is not coding, so getting Rosetta to work with FreeCAD is a personal project that I have not gone too far with. If I do, I will create a code repo. I sometimes code in Python for FreeCAD and Blender3D in the meantime. I like Rosetta, since it is striving to be a general purpose geometry-scripting language like Processing did for multimedia arts. Processing has been ported to, and is key to Rosetta.
There seems to be a lot of ... "not invented here" going on? It seems there's a determination to reinvent or discover stuff that could instead be learned. Like not having familiarity with technical drawing conventions.
Doing it all by trig seems like a lot of work too, a good parametric CAD system would have higher level primitives for curves, chamfer, fillet etc.
Isn't NIH essentially how humans learn? While this particular thing could've been made easier in a mouse-driven CAD tool, and a more pragmatic approach still would be to buy an off-the-shelf tool, this looks like a perfect project for someone who wants to get better at using OpenSCAD.
To an extent, yes. This was a low-risk learning project and I agree that there is value in formally learning the “right way” and not just muscling through (the wrong way, even, sometimes). But this one was just a “quick” side project. I feel like I have a better appreciation for what sorts of stuff I should do upfront.
Also, I am gradually acquiring a sense of space, physical effort, and mechanical operation of things in the real world after a long time in software and feeling like I “couldn’t do” these sorts of things. It’s sort of an eye-opener and a confidence booster to even get through something like this the first time, however sloppily.