In forall x: Calgary, by P. D. Magnus, appears this reasoning:
$\neg(A \land B), A \lor B, A \leftrightarrow B \therefore C$
Examining its truth table, I see there are no rows where all premises happen to be true.
Is this reasoning still valid ?
Yes.
The requirement
In every row where all premises are true, the conclusion must be true as well
can be equivalently reformulated as
There is no row where all premises are true and the conclusion is false
In classical logic, the only way for an inference to become invalid is if there is at least one counter valuation which makes all the premises true but the conclusion false.
If there is no valuation under which the premises are true in the first place, then there can be no such counter example, and the conclusion follows vacuously.
If there is no row that satisfies all of the premises, then this means that the premises are contradictory. In classical logic, according to the principle ex falso quodlibet, from a contradiction anything may be concluded, precisely for the reason that there can be no counter valuation.
First, consider a pair simpler examples:
Example 1
$a\land b \implies b\land a$
This is a valid deduction since it is true regardless of the truth values of $a$ an $b$.
Example 2
$a\lor b\implies a\land b$
This is not a valid deduction since it is sometimes false (on lines 2 and 3 of the truth table).
In the case of your example, it is a valid deduction since it is true regardless of the truth values of $A, B$ and $C$.