Instead of tilting all the rows, I recommend tilting the row or two closer than the home row (possibly more than the buckling spring keyboards do), and then simply raising (without tilting) the rows further than the home row: first, every bit you tilt those further away rows, they get less natural to press (because the switch axis gets further out of alignment with the direction of finger motion); and second, tilting doesn’t actually get you as much height step as you ideally want for the further away rows.
The Model F is an answer to the question, “how do we make something that feels sorta like a beam spring (or selectric), cheaper”. The way IBM answered that question is by dramatically reducing the number of modular per-switch parts, to just barrels, flippies, and keycaps, so that most of the keyboard can be constructed/assembled by robots. Since they wanted to re-use parts, they needed to figure out some way of getting the customary height stagger and slight tilt from row to row, so they angled the keycaps a bit and curved the plate.
But this design only makes sense if you have a single big piece of metal that is easy to bend uniformly. If you have fully separate strips, you can do anything with them, and there’s no reason to copy suboptimal design elements.