I hope this question is appropriate for this site, I was unable to get an answer on Math StackExchange.  

Label $X$ as the real interval $[0, a]$ where $a \in \mathbb{R}^+$, so that $\text{int}(X)$ labels the interior of $X$ and $\partial X$ labels the boundary of $X$.  I have a function $f:\text{int}(X) \rightarrow \text{int}(X)$ that is well defined on the interior of $X$, but ambiguous on the boundary of $X$; that is $f(\partial X)$ is not defined.  

Further, the system has the following properties: $f$ is a contraction mapping over $\text{int}(X)$, and if $a$ is *not* a positive integer, then sequential application of $f$ over any point $x \in \text{int}(X)$ will converge to a fixed point in $\text{int}(X)$.  However if $a \in \mathbb{N}^+$, then application of $f$ over any point $x \in \text{int}(X)$ will converge to a boundary value in $\partial X$.

I want to prove that sequential application of $f$ over $x \in \text{int}(X)$ will converge to a fixed point in $\text{int}(X)$ if $a \in \mathbb{R}^+ \setminus \mathbb{N}^+$, however it is unclear to me if I can use the contraction mapping theorem to achieve this aim.  After all, the contraction mapping theorem requires that the space is complete, however if action of $f$ is only defined over $\text{int}(X)$ then $\text{int}(X)$ is decidedly not a complete space.  

Does anyone know how I can then adequately set up this problem to eventually prove that iteration of $f$ over $\text{int}(X)$ will converge to a fixed point in $\text{int}(X)$ if $a \in \mathbb{R}^+ \setminus \mathbb{N}^+$, and that the limiting behavior of $f$ is undefined otherwise?

Edit: Consider the following Theorem (*Topology*, Gamelin, Green) which will allow me to reword the above in a different way.

**Theorem:** *Let $X$ be a complete metric space with metric $d$, and let $S$ be a metric space.  Let $c$ be a fixed constant with $0 < c < 1$.  Suppose that $(s, x) \rightarrow \Phi_s(x)$ is a continuous function from $S\times X$ to $X$ such that 
$$
d(\Phi_s(x), \Phi_s(y)) \leq cd(x, y)
$$
for $x, y \in X, s \in S$.  Then for each $s \in S$, there is a unique point $x_s^* \in X$ such that $\Phi_s(x_s^*) = x_s^*$.  Furthermore, $x_s^*$ depends continuously on $S$.*

Now, given the above theorem, we can let $a$ be the continuous parameter $s$ as stated in the above theorem, and I am now studying a function $\Phi_a: (0, a) \rightarrow (0, a)$ where the fixed point $x_a^*$ does not depend continuously on $a$, but rather depends *piecewise discontinuously* on $a$, and clearly $(0, a)$ is not a complete metric space as required.

As such, If I was to first establish that iteration of $\Phi_a$ is undefined when $a \in \mathbb{N}^+$, could I then apply the above theorem for all $a \in \mathbb{R}^+ \setminus \mathbb{N}^+$ and put $\Phi_a: [0, a] \rightarrow [0, a]$ so that we have a complete metric space and the behavior of the map is well defined for all such values of $a$?