Urban localization with the Global Navigation Satellite System (GNSS) remains an unsolved problem with frequent misdetection of which street or which side of the street the user is located. 3D mapping-aided GNSS uses grid-based GNSS shadow matching and ranging alongside data-driven line-of-sight (LOS) classifiers to improve localization accuracy. However, previous work on shadow matching has not considered the needs of risk-aware autonomous systems. Our prior work introduced a fully set-based version of shadow matching that proposed set-valued maps and localization with computationally efficient set representations. In this work, we extend our prior work to propose Mosaic Zonotope Shadow Matching (MZSM), which provides a mathematical framework to begin addressing risk awareness in 3D mapping-aided GNSS. We employ a classifier-agnostic polytope mosaic architecture to provide risk-aware set-based bounds on urban positioning. We formulate a recursively expanding binary tree that, through set operations, refines an initial set-based location estimate into a mosaic of smaller polytopes. We assess our framework with a 3D building map of San Francisco and emulated classifiers to validate our algorithm's risk-aware improvements. We demonstrate that MZSM provides set-based localization bounds across varied data-driven LOS classifiers and yields a more precise understanding of the uncertainty and risks in 3DMA-GNSS over existing methods. We validate that our tree-based construction is efficient and tractable, computing a mosaic from 14 satellites in 0.63 seconds and growing quadratically in the satellite number.