Another question about absolute error

Hi again...
It is known that the absulute error of the quotient or product of a number of quantities is less than or equal to the sum of their absolute errors...(1)
and
Absolute error = Observed - Accepted value (2)

but absolute error can be < 0 in some of the quantities of the product or the quotient (if in (2), Observed < Accepted) and in some of the quantities it can be >= 0....if I folow (1) I have to sum the separate absolute errors but actually they can be for example:
-12+14+(-2), it is equal to 0...so is it true that the absulute error of the quotient or product is <= 0, and shall I accept 0 as the most possible absolute error of the quotient or product......or I have to use
|absolute error| every time when I compute it and in this way it is >=0 every time!!!
Thanks all!