I find that the typical arm parts allow for more flexible power handling. For example, you have control over the core frequency and have separate control over each peripheral clock. So you can really fine tune as needed.

Most parts are quite capable of 8051 type <1uA sleep current. However, it takes more work (time) to minimize overall power consumption. You have more peripherals, more I/O (generally), and more bus structures to tweak....ergo the additional time spent.