There are different implementations below the covers on these two methods right now, and essentially backed by different kinds of caches. Their results will become more aligned within a few weeks. Most numbers in the Twitter API are “eventually consistent” anyway – lots of servers to connect to, lots of chances for varying values.