Simulation results are presented for all test cases prescribed in the Second AIAA Sonic Boom Prediction Workshop. An inviscid, embedded-boundary Cartesian-mesh flow solver is used to compute "boom carpets"-pressure signatures at a range of specified distances and off-track angles. To accelerate the process, each boom carpet is automatically decomposed into several independent meshes, computed in parallel. Each mesh uses output-based mesh adaptation to affordably obtain credible results. This paper discusses improved Cartesian meshing strategies for boom prediction, including Mach alignment, azimuthal alignment, high-aspect-ratio cells, and adaptation functional weighting. The resulting nearfield signatures generally exhibit good convergence with mesh refinement. For an independent evaluation of our results, this study introduces a local error estimation procedure that highlights regions of the signatures most sensitive to mesh refinement. Results are also presented for the two atmospheric propagation test cases, which investigate the effects of atmospheric profiles on ground noise. Propagation is handled with an augmented Burgers' equation method (NASA's sBOOM), and ground noise metrics are compared.