I have recently read a paper which stated that "by using tandem-MS as a specific detector, poor peak shapes are acceptable.". I am wondering why and wether the poor peak shapes affect the assay accuracy or not. Thank you for any input.

Peak shapes should always be good so that they can be consistently integrated. They just don't have to be resolved if MS is used. I like to keep mine resolved, though. This is because 1) people like to see resolved peaks and it leaves me less explaining to do (e.g. to boss and clients), and 2) the effects of matrix in MS and MS/MS are often underestimated, and they can be lessened if the peak elutes by itself.