Assuming Reals, then including Assumptions → x1 > 0 && x2 > 0 in this Limit would seem to be necessary in order for f to have defined values for any n.
Even with this amendment though, the result is not as expected (i.e., Max[x1, x2], or something equivalent, because by symmetry, if the limit is x2 when x1 < x2, then the limit is x1 when x2 < x1).
Since this isn't happening, I think you've found a gap in Limit's coverage.