The use of soft, compliant actuators has recently gained research attention as a potential approach to improve human-robot interaction compatibility. Fluidic artificial muscles, or McKibben actuators, are a popular class of soft actuator due to their low cost and high force-to-weight ratio. However, traditional McKibben actuators face efficiency problems, as in most actuation schemes, the actuator is sized for the largest possible load, resulting in energy loss when operating at lower force regimes. To address this issue, our group has developed a bio-inspired actuation strategy called variable recruitment. In variable recruitment, actuators are placed within a bundle and can be sequentially activated depending on the required load. This strategy mimics the hierarchical architecture of mammalian muscle tissue and improves system efficiency and bandwidth while allowing for variable stiffness properties. Previous variable recruitment models and controllers assume that the force output of each actuator is independent and that these forces sum to provide the total bundle force. However, our recent work has shown that there is significant interaction between actuators within a bundle, particularly at lower recruitment states. This is because at these states, inactive or partially activated actuators resist bundle motion and reduce total force production. In this paper, we study these resistive effects at low recruitment states by considering two different variable recruitment configurations: a fixed-end configuration (with resistive forces) and a tendon configuration (designed with tendons to eliminate resistive forces). We then assess the tradeoffs between the two configurations. We found that while using the tendon configuration eliminates resistive forces, if we consider both configurations with the same overall system length, the tendon configuration has less overall system free strain because its FAMs have to be shorter than those of the fixed-end configuration. However, despite this difference in free strain, our results still show that the tendon configuration can have higher maximum load capacity and efficiency than the fixed-end configuration and that the specific application and system requirements will dictate the proper configuration choice.