In the book of Analysis on Manifolds by Munkres, at page 137, it is given that
I have doubts about the existence of such a $C_x$, so just to make sure, I wanted the prove the existence of such a set $C_x$.
Proof:
Let assume that for some $x_0 \in B_i$, every neighbourhood $C_{x_0}$ of $x_0$ has a non-empty intersection with $D_{i-2}$. Then since $B_i \subseteq Ext(D_{i-2})$, and $C_{x_0} \cap B_i \not = \emptyset$, then it must be true that $$x_0 \in \partial D_{i-2} \subseteq D_{i-2},$$ but $B_i$ and $D_{i-2}$ are disjoint, a contradiction.
Is there any flaw in the proof ? I'm particularly asking this because it took me 1 day to figure out this contradiction out clearly.

I don't feel like reading your proof, but here's the proof that comes to my mind. $B_i$ and $D_{i-2}$ are disjoint compact sets, so they are some positive distance apart from each other. So given any $x \in B_i$, we can take a small enough cube centered at $x$ that is disjoint from $D_{i-2}$ (we can insist the cube is contained in $A$, since $A$ is open).