The touch screen was always going to be a bit of gamble I feel. I wonder if the stick/throttle controls can interact with it for more basic commands?
They can; one of the hats on the throttle lets you slew around a cursor, though I don't know how that works across various portals on the display - I would assume you can't slew the cursor from one side of the screen all the way to the other, and that it wouldn't be available on every page, but that the pilot would have to use another hat to toggle something like a sensor (or portal) of interest. Maybe I'm wrong though and the cursor can be used anywhere on the display.
I may not be an F-35 pilot, but I think moving away from touch screens would be regression - instead I think they need to better leverage modern consumer and commercial touch screen tech to incorporate things like capacitive touch layers (with IR as a back-up in case chemical warfare agents or water / moisture screws with the capacitive functionality), as well as the algorithms that've been developed to produce things like palm rejection in good modern tablets / touchscreen laptops. Other things such as predictive / speculative object selection could also be trialed / investigated, where if a pilot's finger misses (eg) a virtual button, but there are indications that the touch wasn't accurate (such as the finger / cursor rapidly sliding across the screen) then an algorithm can select the button they were closest to touching. Or if the finger has a lot of erratic movement around the screen (perhaps with accelerometer data indicating turbulence), in an area where there are multiple virtual buttons or targets on a display, the algorithm can go ahead and select the one that statistically (or based on the F-35's combat ID systems and threat prioritisation algorithms) that's more important. For example, if there's a hostile and a friendly close to each other on a tactical display, there's indications of turbulence / buffeting, and the pilot's finger has hit the screen closer to the friendly track, the display might instead select the hostile track.
I know such a system would have plenty of potential for error, but if you combine it with a physical manual override switch (just use one of the ~20 hats and switches on the HOTAS with the right context, or if you really want, add a new physical switch), I think it'd be sufficiently safe. Have the software written by someone competent with a reasonable fixed-cost contract (the concept / tech isn't cutting edge in the consumer world) and I think it could surprise people with how accurate it is (a number of consumer touchscreen devices already do this; the word prediction on good phone keyboards like Swiftkey (owned by Microsoft) can do an amazing job despite the open-ended problem it faces (but which the F-35's display doesn't)).