Since you were mentioning truth tables, I asssume you're talking about classical propositional logic, and one of the standard natural deduction calculi for it.
Then, your first conjecture that
($\Gamma_2 \vdash B_2$ is provable) $\Rightarrow$ ($\Gamma_1, \Gamma_2 \models B_2$ is true)
is correct. By soundness of natural deduction, we have
($\Gamma_2 \vdash B_2$ is provable) $\Rightarrow$ ($\Gamma_2 \models B_2$ is true)
By monotonicity of classical logic, we have
($\Gamma_2 \models B_2$ is true) $\Rightarrow$ ($\Gamma_1, \Gamma_2 \models B_2$ is true)
Application of transitivity yields the desired result.
Your second conjecture that
($\Gamma_1, \Gamma_2 \models B_2$ is true) $\Rightarrow$ ($\Gamma_2 \vdash B_2$ is provable)
is not correct, however. Here is a counterexample: Set $\Gamma_1 = \{P\}$, $\Gamma_2 = \{P \rightarrow Q\}$ and $B_2 = Q$. Then $\Gamma_1, \Gamma_2 \models B_2$ is true (check with a truth table, if you like). But $\Gamma_2 \models B_2$ is not true, so by soundness again $\Gamma_2 \vdash B_2$ is also not provable.
Nonetheless (if that should be what you've meant),
($\Gamma_2 \models B_2$ is true) $\Rightarrow$ ($\Gamma_2 \vdash B_2$ is provable)
is correct, because natural deduction is also complete.