Think of your agenda as a large structure with different slots that can be filled with particular tasks, large and small. Some ideas will develop into more significant contributions to the agenda; others will be lost because they don't support the agenda, and as a consequence aren't supported by it.
You can think of your overall research agenda on the model of the hermeneutic circle. The agenda gives meaning to the individual parts of it (chapters or articles), and the individual parts of it modify the overall definition of the project. For example, I might write a chapter that ends up substantially modifying my conception of the entire book, forcing other chapters or ideas for chapters out.
This model is preferable to the idea of getting an idea for a single piece, and then trying to complete that piece. If you don't know what it's a piece of, it's harder to write.