I'm not following your argument well enough to see exactly where it goes wrong... One question you might ask yourself is "does this show that every closed set is countable?" What is special about the cantor set here? I'm not seeing it.
As for why the cantor set is uncountable, consider this:
At each finite level of the cantor set construction, we "throw out" the middle third of each piece. So we have a decision to make at each stage: Do we go left? or do we go right?
E.g., We start in $[0,1]$. Then We have to decide to go into $[0,\frac{1}{3}]$ or into $[\frac{2}{3},1]$. Let's say we go left. Now we have the choice of going into $[0,\frac{1}{9}]$ or $[\frac{2}{9},\frac{1}{3}]$.
You can see that every countable sequence of choices (left or right) gives a unique point of the cantor set. Moreover, every point of the cantor set corresponds to such a sequence of choices. So if we write $0$ for "left" and $1$ for "right, the points of cantor set are in bijection with the infinite strings of $0$s and $1$s.
As a fun aside, the topological structure actually agrees as well! That's why you'll often see people call the cantor set $2^\omega$. In set theoretic language, that basically translates to "infinite sequences of $0$s and $1$s".
Ok, but now there must be uncountably many infinite sequences of $0$s and $1$s by a diagonalization argument. So the cantor set is uncountable too.
I hope this helps ^_^