Due to the various stakes, values and views of social groups involved with climate change and adaptation, the process of developing the Dutch National Adaptation Strategy (NAS) needs to take a plurality of frames into account. The PBL Netherlands Environmental Assessment Agency (PBL) aimed to inform this process using frame analysis. However, researchers at PBL did not succeed in applying the method as planned. Over the course of the production of the NAS, the hegemonic science-risk frame, which focuses on quantitative identification and subsequent prevention of risks, emerged as the dominant frame. Our case analysis based on participant observation and interviews shows that, even when frame-reflection was explicitly aimed for, this happened to be downscaled, unwittingly, under influence of tensions, challenges and paradoxes encountered during the essential balancing act that characterizes complex science-policy interfaces. Roles, interaction processes, client needs, internal processes are dynamically shaping and shaped by institutionalised expectations over objectivity, independence, inclusiveness and effectiveness. We argue that what makes frame analysis worthwhile is not so much its presupposed power to lead to the adoption of a multiplicity of frames, but rather its ability to lead to a form of institutionalized critique that refuses to take automatic recourse to a dominant frame (e.g. the science-risk frame). Thus, frame analysis is a crucial instrument in performing the aforementioned craft of science-policy interfacing, and needs to be more firmly integrated into science-policy interfacing for this very reason.