Good question. You may not need to worry about outliers when doing 1:1 matching since units without a similar pair will be ignored.

In other cases, outliers can "skew" the analysis, however I (personally) am always hesitant to throw out data unless I am very confident the dropped records are not relevant to the analysis.