Let $\mathcal{L}$ be the formal system of statement calculus. Let $A$ be a formula of $\mathcal{L}$. Then each of the three formulas
\begin{gather*}
A \vee (\sim A) \\
A \implies [(\sim A) \implies A] \\
(\sim A) \implies [A \implies (\sim A)] \\
\end{gather*}
is a theorem of $\mathcal{L}$.
By case analysis,
$$[(\sim A) \implies A] \vee [A \implies (\sim A)]$$
is a theorem of $\mathcal{L}$. But this seems intuitively incorrect.
For example, each of the statements
\begin{gather*}
(\forall x \in \mathbb{N})[(x < 3) \implies (x \ge 3)]\\
(\forall x \in \mathbb{N})[(x \ge 3) \implies (x < 3)]
\end{gather*}
is false. However, the disjunction of these two statements is true, by the argument above.
Edit. I guess the example was not very good.
Suppose you flip a coin. It is a fair coin and it cannot stand up on its side. It lands at your feet on a solid, flat surface.
Let $A$ be the statement that the coin shows heads. Then $(\sim A)$ is the statement that the coin shows tails. The statement
\begin{gather*}
\text{"if the coin shows heads, then it shows tails,}\\
\text{or}\\
\text{if the coin shows tails, then it shows heads"}
\end{gather*}
is true, even though both disjuncts are false.
Maybe this question no longer belongs on math.stackexchange.com
.
Best Answer
$$ (\forall x \in \mathbb{N})[x < 3 \implies x \ge 3] \lor (\forall x \in \mathbb{N})[x \ge 3 \implies x < 3] $$ is not true, and your analysis doesn't imply that it ought to be.
What is true by your argument is $$ (\forall x \in \mathbb{N})\bigl[(x < 3 \implies x \ge 3) \lor (x \ge 3 \implies x < 3)\bigr] $$ but you can't move the quantifiers into the middle of its structure and expect it to remain true.
In general, $(A\to B)\lor(B\to A)$ is always true in classical logic, no matter whether $A$ and $B$ are negations of each other or not. Perhaps even more intuition-challenging, $(A\to B)\lor(B\to C)$ is always true too.
Response to edit: I think you're being confused by two different uses of "if ... then" in mathematical English.
On one hand in casual non-formal mathematical English, "if it's heads, then it's tails" would mean something like
(and one then hopes that the context makes it clear which situations are "relevant").
On the other hand, in (classical) formal logic, the formula "${\rm heads}\Rightarrow{\rm tails}$" means
Which is the same as saying
Confusion creeps in because "${\rm heads}\Rightarrow{\rm tails}$" is often pronounced "if heads then tails", as if its meaning were the same as in (1), but it is not when formulas are involved.
The sense in which you can prove that $(A\Rightarrow \neg A)\lor(\neg A\Rightarrow A)$ is true is (2) and only (2). It stops being true if you try to reinterpret the $\Rightarrow$s as if they meant (1).
Furthermore it it looks like you're also confusing yourself by the difference between "is true" and "has a proof" when you say that "this or that" is true but "this" and "that" are both false.
What you mean here is that you believe you have a proof of "this or that". But clearly it is not true because the definition of "or" is that "this or that" is true exactly when at least one of "this" and "that" is true -- and you have just argued that they are both false.
So when you ask,
the answer is that they can't and you're mistaken about at least one of those parts. Same as if you have an apparent proof that concludes $2+2=3$: you shouldn't be asking "why is $2+2$ not $4$?" but "what is wrong with this thing that looks like a proof but obviously can't be?"