That cell was the only formula -- all other cells were static amounts with no more than two places. Huh? 2/2
@johns seriously, the way it rounds numbers (internally binary floats, which is a design mistake) to decimals for display just doesn't agree with how anyone else, including the standard defining what numbers are meant with this binary representation, rounds things.
Financial math of course rounds differently than IEEE754, but it's also not what oocalc does, because that would require internal math to be done with decimals and fixed precision and not binary floats getting fin-rounded afterwards!
@johns I did an experiment on that a way back. Libreoffice calc does a weird hybrid where it tries to adhere to financial math rounding, but relies on IEEE754 double precision (machine) floating point rounding. It has its own digit-rounding way to display numbers (which is different to the math it does below); I came to the conclusion it's broken by design. You might have hit one of the points where the display rounding is correct but the underlying numbers are incorrectly truncated by formulas