The problem is of incompatible time scales. The original studies and requirements for the F-22 were done in the 1980s, Lockheed selected to build the F-22 in the 90s, F-22 entered flight test in 1997, IOC declared in 2005, expected to be in service until the 2040s at least.
Iteration time goes down by a lot in wartime. Also, choosing technologies which have much shorter iteration times could be a disruptive game changer. I'd bet shaped charge carrying drones with a range of 4 km would be much faster to iterate on than tank guns with the same range.
If I were Taiwan, I would be working with the US to engage in a crash program of developing AI drone mini-subs. Basically, develop the ability to make the waters around Taiwan exceedingly dangerous, even without air superiority.
Iteration time goes down by a lot in wartime. Also, choosing technologies which have much shorter iteration times could be a disruptive game changer. I'd bet shaped charge carrying drones with a range of 4 km would be much faster to iterate on than tank guns with the same range.