I am not entirely sure what your issue is, I will address what I think the possible issues are.
The iteration is defined by $x_{n+1} = x_n - { f(x_n) \over f'(x_n) }$.
You can do computations in $\mathbb{R}$ or $\mathbb{C}$. In either case, you must have $f'(x_n) \neq 0$ for the iteration to be well-defined.
If the function maps reals to reals (that is, $f(\mathbb{R}) \subset \mathbb{R}$), then if $x_0 \in \mathbb{R}$, and the iterates are well-defined, then the above equation shows that $x_n \in \mathbb{R}$ for all $n$. So, $x_n$ can never 'turn' complex using this scheme.
If you are doing computations in $\mathbb{C}$, then it is entirely possible that the iterates will be complex.
It is entirely possible that the system $f(x) = 0$ has no solutions in either the real or complex numbers. For example, take $f(x) = 1$.
Here is a more illustrative example:
Let $f(x) = x^2+1$. Note that if $x$ is real, then $f(x)$ will be real.
As a function $f: \mathbb{C} \to \mathbb{C}$, the equation $f(x)=0$ has two solutions ($\pm i$, of course).
As a function $f: \mathbb{R} \to \mathbb{R}$, the equation $f(x)=0$ has no
solutions (since $f(x) \ge 1$ for all real $x$).
If you start with $x_0$ 'near' $\pm i$, and use complex arithmetic, then $x_n$ will converge very quickly to $\pm i$.
If you start with $x_0 \in \mathbb{R}$, then either $f'(x_n) = 0$ for some $n$ (in which case we must have $x_n = 0$, and subsequent iterates are not defined) or the iterates do not converge (if they did we would get a contradiction for this function). However, if they are defined, then the iterates are all real.