For instance, they could make F2 something that was infrequently useful and unintrusive, then made F3 do what F2 did. So if you mispress, you don't do the even worse F1 thing.
F2 has long been "Rename Object" and F3 has long been "Find" (though Excel in this case doesn't use it for that, but that's another matter). Did IBM do enough user studies at the time? I'm not sure IBM could have found a user study to predict the order of those choices being irritating to Excel users two and half decades later.
Even then, too, the IBM CUA I pointed to was partly synthetic, pulling in commonalities from apps at the time of its writing. Given F2 is nearly universal as "Rename Object" [1], I've got a feeling it predates the IBM CUA by quite a bit in some heavily used applications, but I'm not a historian.
Office never existed in a vacuum, and the function keys have had common meanings for decades. Microsoft used a mix of the IBM CUA guidelines and presumably what they were seeing from the competition (Lotus 1-2-3 and VisiCalc especially for Excel early in its history). If they put the useful function of F2 somewhere else, they would have broken the muscle memory of so many other PC users, even early in Excel's history. If they moved it today there might be riots, some rioting even by the same people removing F1 keys from keyboards.
[1] You can see it on this chart and compare it with most of its neighbors: https://en.wikipedia.org/wiki/Table_of_keyboard_shortcuts
So it is very hard to trade-off the momentum of a three decades old standard just to delight particular subsets of power users.
Not to mention the other side of it: why was help so slow and aggravating? Were there ways to make it non-modal and faster. Unsurprisingly, that's been the focus where Office has tried to make an impact, trying to speed up Help and make it less obnoxious over the years, rather than change up a three decade old keyboard shortcut.