The ‘Problem of Unconceived Alternatives’ – essentially the idea that we can never know when a radically different but better explanation is available – goes to the heart of what is involved in trying to understand the cosmos given our limited capacities for observation, and the challenges of interpreting the data. This article rethinks large-scale cosmological interpretation (in effect, ‘metaphysics’) as a process of modelling ‘protectorates’ of past experience in terms of ‘typicalities’ found in our own local range of empirical data, and then of making it available as a tool for understanding and prediction. Based on the role of examples and analogies (dṛṣṭānta) to build ontologies explaining the cosmos in the history of Indian metaphysics, it argues for a broadly structural realist account. When we ask whether something is a physical object, a material, a force, a field, or some other as-yet-unconceived kind of thing, we use best-fit models that are schematic of the structure of evidence, rather than descriptive of the thing in itself. Given this, Indian metaphysical history suggests strategies for finding unconceived alternative better explanatory models, by stretching the imagination towards novel schemas. In this light, the ‘problem’ becomes a ‘promise’ that unconceived alternatives with ever-better explanatory power await us, subject to more innovative, imaginative interpretations.