If two experiments measuring the same quantity get 2±1 and 10±1, the final result isn't something like 6±0.7, it's something like 6±4.
@awsteiner the errors don't add like resistors in parallel ;)