Many mutualisms are taken advantage of by organisms that take rewards from their partners but provide no benefit in return. In the absence of traits that limit exploitation, facultative exploiters (partners that can either exploit or cooperate) are widely predicted by mutualism theory to choose an exploitative strategy, potentially threatening mutualism stability. However, it is unknown whether facultative exploiters choose to exploit, and, if so, make this choice because it is the most beneficial strategy for them. We explored these questions in a subalpine plant-insect community in which individuals of several bumble bee species visit flowers both "legitimately" (entering via the flower opening, picking up and depositing pollen, and hence behaving mutualistically) and via nectar robbing (creating holes through corollas or using an existing hole, bypassing stigmas and anthers). We applied foraging theory to (1) quantify handling costs, benefits and foraging efficiencies incurred by three bumble bee species as they visited flowers legitimately or robbed nectar in cage experiments, and (2) determine whether these efficiencies matched the food handling tactics these bee species employed in the field. Relative efficiencies of legitimate and robbing tactics depended on the combination of bee and plant species. In some cases (Bombus mixtus visiting Corydalis caseana or Mertensia ciliata), the robbing tactic permitted more efficient nectar removal. As both mutualism and foraging theory would predict, in the field, B. mixtus visiting C. caseana were observed more frequently robbing than foraging legitimately. However, for Bombus flavifrons visiting M. ciliata, the expectation from mutualism theory did not hold: legitimate visitation was the more efficient tactic. Legitimate visitation to M. ciliata was in fact more frequently observed in free-flying B. flavifrons. Free-flying B. mixtus also frequently visited M. ciliata flowers legitimately. This may reflect lower nectar volumes in robbed than unrobbed flowers in the field. These results suggest that a foraging ecology perspective is informative to the choice of tactics facultative exploiters make. In contrast, the simple expectation that exploiters should always have an advantage, and hence could threaten mutualism persistence unless they are deterred or punished, may not be broadly applicable.