No, not in general.

My metric space is the disjoint union of uncountably many copies of $\mathbb R$.
$$X = \bigsqcup_{t \in T}  X_t$$
where $T$ is uncountable and $X_t = \mathbb R$ for all $t$.  The metric: two points in the same $X_t$ have the usual distance, two points in different $X_t$ have distance $1$.  
My measure is Lebesgue measure $\mu_t$ on each copy $X_t$ of $\mathbb R$.  So for a subset $E \subseteq X$ we can write $E = \bigsqcup_{t \in T} E_t$ where $E_t \subseteq \mathbb R$, and its measure is $$\mu(E) = \sum_{t \in T}\mu_t(E_t).$$  

This measure is locally finite.  Any point in $X$ lies in exactly one set $X_t$ and the open ball of radius $1/2$ centered there has measure $1$.

But your finiteness property fails.  Let the closed set be $$F = \bigsqcup_{t \in T} F_t$$ where $F_t = \{0\}$ for all $t$.  Then $\mu(F) = 0$.  Let $G \supseteq F$ be an open set.  I claim $\mu(G) = +\infty$.  Indeed, $$G = \bigsqcup_{t \in T} G_t$$ where for all $t$, the set $G_t$ is an open neighborhood of $\{0\}$.  So $\mu(G_t) > 0$ for all $t$.  And $\mu(G) = \sum \mu_t(G_t)$ is an uncountable sum of positive numbers.  So $\mu(G) = +\infty$.