I'm using mvernacc's Aerospike Nozzle Design GUI, which is supposed to be able to design and predict the performance of aerospike rocket engines. The workings of the software are apparently based on this paper. While this is the software I'm using, the question is more about aerospikes in general, which the software seems to show behaving unexpectedly, rather than the specifics of using the software, hence asking it here rather than in the GitHub issues for that project.
I've loaded up the default example file, and raised the expansion ratio (ratio of lip area to throat area) from 12 to 60. In this configuration, it predicts a rather poor sea level Isp of 85.2 seconds. Decreasing the expansion ratio to 8, the Isp is 206 seconds. Meanwhile, in vacuum, it predicts an Isp of 250 seconds at an expansion ratio of 60, and an expansion ratio of 230 seconds at an expansion ratio of 8.
I understand that even with an aerospike, performance in vacuum should be better than performance at sea level, but these results go against my understanding of how aerospikes work. If aerospikes are inherently altitude-compensating, why is a good expansion ratio for sea level performance seemingly so far from a good expansion ratio for vacuum performance? Shouldn't performance at any given pressure be equal, no matter the throat to lip area ratio? Isn't choosing an expansion ratio for your design altitude the exact problem altitude-compensating nozzles are supposed to avoid?
Am I misunderstanding aerospikes? Does altitude still matter and simply matter less than it would on a converging-diverging nozzle, or is the software wrong, or am I doing something wrong?