The goal for this research is to articulate and test a new hybrid Augmented Reality (AR) environment for conceptual understanding. From the theoretical lens of embodied interaction, we have designed a multi-user participatory simulation called ARfract where visitors in a science museum can learn about complex scientific concepts on the refraction of light through full-body immersion using optical see-through AR glasses, projection-based AR, and gesture technology. In particular, we developed two different types of simulations for ARfract, namely a "game-based simulation" and a "non-game simulation" to explore how the order of different AR simulations influences the perceived usability, user behaviors, learning experiences, and learning outcomes. For the experiment, 10 dyads were randomly assigned to one of the two experimental conditions: 1) the game-to-non-game condition and 2) the non-game-to-game condition. The results indicate that the learners who experienced the game-based simulation before the non-game simulation performed better than did the other group with the reversed experience order. This paper also reports the usability, user behaviors, and learning experience issues regarding the affordances of hybrid AR technologies. The major contribution of this proof-of concept research is that it articulates our understanding of how particular configurations (i.e., order) of the emerging technologies (i.e., hybrid Augmented Reality systems) and its use can lead to different learning outcomes.