Let $A$ be a semilocal Noetherian ring with Jacobson radical $m$ and $M$ a finite $A$-module. Let $x \in m$. According to Matsumura's Commutative Ring Theory p. 99 (Step 2), $l(xM/xM\cap m^n M)=l(M/(m^nM:x))$. It seems to me that this equality is not true via the following argument:
First note that $xM \cap m^n M = x(m^nM:x)$. Let $f$ be the composite map $M \rightarrow x M \rightarrow xM/xM\cap m^n M \rightarrow 0$, where the first arrow is multiplication by $x$. We can rewrite $f$ as $M \rightarrow x M \rightarrow xM/x(m^n M : x) \rightarrow 0$. Now, $(m^nM:x) \subset Kerf$, but since $M$ is not in general torsion free, we have an exact sequence $0 \rightarrow K \rightarrow M/(m^nM:x) \rightarrow xM/x(m^nM :x) \rightarrow 0$, with $K$ nonzero in general. Since the length is additive function on short exact sequences, we are done.
Question: Any comments on the above? Am i missing something or is there a typo in Matsumura's book?