I've not looked at the second order expansion yet, but the problem with the first example is that we've got a recurrence relation for the derivative of struve_h(v, z) in terms of struve_h(v-1, z). When taylor tries to evaluate the derivative at z=0, it has to evaluate struve_h(-2, 0), which is a pole.

Of course, struve_h(-1, x) is analytic so this isn't particularly impressive...