First, let me apologize for my post that triggered your outburst!
I am tempted to say that I don't know what triggered it, but I' be lying. I know exactly what triggered it!
Lets keep it at that and get back to the topic.
The initial title speaks about auto dimensioning not sheet metal bending, but never mind. I stick to what I said in my first reply. There is a reason that I said what I said and it based on more than 3 decades of experience in the field. Auto dimensioning is only "really" possible when the design intent is captured in the sketches and the way they are dimensioned in the sketches and in the design history. That capture of design intent is often different from manufacturing method and dimensioning a drawing is usually a mixture if trying to convey design intend and manufacturing methods.
A STEP file which only contains BRep data, meshes or surfaces but does not contain any design history, only geometry. You cannot judge the design intent and the manufacturing intent purely by looking at geometry, which is the end result of a design process. Heck most CAD systems do not even manage decent feature recognition in more complex geometry.
This is not a deterministic problem that can be solved with a simple algorithm. One day AI and builtin expert knowledge will surely be able to help that aspect but until then we're stuck with what we have. In Fusion 360 you can design fine in direct modeling mode without a timeline. Sketches are not associative with the geometry thus would be dangerous to use for data used in auto dimensioning.
To close this chapter, I've worked for a company that had over 200 seats of Solid Works. I don't know a single engineer that really truly enjoys dimensioning drawings. In fact most hate and despise it! Thus a lot of brain effort has gone into research of how it can be accomplished. If this would be trivial it'd been solved 20 years ago.
Let's move on to the sheets metal problem you are encountering. Welcome to the world of small volume manufacturing!!!
In high volume manufacturing usually lot of data is recorded, stored and evaluated, often in real time to see if the manufacturing processes are in statistical process control.
Tooling is very purpose built, qualified for and troughout production through stringent qualification checks against known, calibrated standards.
Tooling is designed using known design standards and formulae which often are good approximations of real world behavior but still are just approximations. Approximations by nature assume a certain error and often there are enough variables that errors accumulate or even magnify one another. The end result is that the manufactured tooling does not produce parts/product to specifications. Experienced engineers, technicians and machinists usually have an intuitive grasp of what to do to correct the issue. Tooling is modified based on guidance by these experienced people until the product does match the specifications and then drawings and documentation are adapted to reflect tooling reality ( not the other way around !!! ) so a new tool can be manufactured to reflect these realities and product matching specifications can be produced directly off the tooling without having to go through this often lengthy modification process.
The K-Factor is one of these engineering approximations that is used to predict the outcome based on assumptions of actual manufacturing conditions.
It's a fairly simple rule of thumb that works but not always. If your sheetmetal shop is not able to match these precise conditions as it would be too expensive to build purpose build tooling for a one-off or even a small series then these approximations simply fail.
You should not be mad at these sheetmetal shops, you should be happy that they tell you. Those are the guys I'd go back to, because they know what they are doing!
Finally, there is absolutely nothing wrong with being a noob. We all have to start somewhere!
If one want's to really learn, it is imperative to listen to others and accept their obvious expertise because otherwise one will remain a noob!