I've tried with and without tables using all 3 formulas and a variety of cell formats, for me the results are consistent with all of them, either value or error depending on the content of the cell, but never a different result for one formula.
As I said in post 5, this is something that will be practically impossible to diagnose without the original data.