Yes. That is a kind of first iteration of tool that is far to be solid.
Settings that are displayed in last operator panel are the ones exposed when adding primitives using Add menu (shift A).
So, yes. Adding a cube or a sphere just displays a unique radius setting.
Adding a Cylinder displays a radius and a depth setting.
Adding a Cone displays 2 radius and a depth setting.
So, we are limited to those primitives.
And when you try to snap a primitive on a rotated object, you will encounter issues.
Frankly, that tool need work to be a real equivalent to what exists in other software.
People are thinking that ease of use is as easy to create. But that is the opposite and that takes time.
With 2.8 design, Blender is trying to assimilate ease of use of tools well-known in other software.
But this is a target for a frightening amount of tasks.
So, developers choose to attack the mountain by completing a first milestone for each task instead of letting people think that they will not intervene.
They are currently developing Geometry Nodes. We should end up with some kind of parametric primitives using those nodes. In few years, we should end-up with a smooth workflow when using Add Object tool could create a parametric primitive.
But that will take years because a lot aspects of Blender are in the same status of half achieved feature, first milestone.
Campbell literally spend months working on current Add Object tool. Before that, he spend month solving problems with the Scale Cage active tool that did not exist before 2.80.
You can use this tool and keep an eye on dimensions in Transform panel to modify individual dimensions of object.
You have to precise that such vertex is the point of reference for rotation.
Whatever software you are using, you are doing something similar.
You can be in a default mode that automatically choose pertinent vertex close to mouse pointer.
But in that case, if you want to rotate around an object origin instead of a vertex, you have to quit that mode and do operations to change pivot point.
That was worst in old Blender. You had to go into edit mode, to change select mode to select vertex, to select vertex, to use Shift S menu to snap 3d Cursor to vertex, to quit edit mode.
Now, using 3D Cursor active tool, that is quicker to snap 3D Cursor on a vertex, without quitting Object mode.
What you are used to is not very much faster. I concede that can be faster but not very much faster.
You probably have 2/3 operation to do, anyways.
I think that nothing can be qualified as natural. You are just used to that way.
People using another software could be annoyed trying the one you are used to.
And about versatility, you would have to be more explicit.
In Blender, you have several choices.
You can use center of bounding box of selection, origin of object, individual origins of all objects of selection, median point from location of origins of objects of selection, active element as a pivot point.
Or you can use any location as a pivot point using 3D Cursor.
You can be annoyed about slow way to place 3D Cursor.
But that is simply impossible to be more versatile about placement of pivot point than being able to place it, anywhere.