I would be happy with just knowing the rms error delivered by one spreadsheet as compared to another.This is exactly why I'm hesitant to make changes. The regression testing needed to confirm the accuracy of predicted to real-world pH values is significant (think years).
Also keep in mind that if you do something based on what I have posted in this thread (augmented with places for additional grains) you get a detailed look at where all the protons are going. You can take the data from an old brew and plug it into your spreadsheet and this one and see where they differ and what may be driving the error you got. For example you may see in a brew whose actual pH was appreciably lower than predicted that one malt was a particularly strong proton absorber and re-examine your model for that malt. Or you may see that your program requires more acid for water of a given alkalinity than this one does and focus on how you handle alkalinity calculations.
My first thought was that the main value of the spreadsheet offered here was for teaching but it has since occurred to me that it has potential value as a spreadsheet developer's diagnostic tool as well.