Your algorithm is incorrect. I assume you know how to compute the suffix array and the LCP array of a string, that is, their efficient implementation. As has been pointed out in the comments, you should try to understand what each component is, and why it works.
First of all, is the suffix array ($SA$) of a string. A suffix array is basically all the suffixes of the string $S$ arranged in ascending lexicographic order. More specifically, the value $SA[i]$ indicates that the suffix of $S$ starting from position $SA[i]$ is ranked $i$ in the lexicographic ordering of all suffixes of $S$.
Next is the $LCP$ array. $LCP[i]$ indicates the length of the longest common prefix between the suffixes starting from $SA[i-1]$ and $SA[i]$. That is, it keeps track of the length of the longest common prefix among two consecutive suffixes of $S$ when arranged in lexicographic order.
As an example, consider the string $S = abbabca$. The suffixes in lexicographic order would be $\{a, abbabca, abca, babca, bbabca, bca, ca\}$, so $SA = [7, 1, 4, 3, 2, 5, 6]$ for a 1-indexed array. The $LCP$ array would be $LCP = [-, 1, 2, 0, 1, 1, 0]$.
Now, given two strings $A$ and $B$, we concatenate them as $S = A\#B$, where $\#$ is a character not present in both $A$ and $B$. The reason for choosing such a character is so that when computing the LCP of two suffixes, say $ab\#dabd$ and $abd$, the comparison will break off at the end of the first string (since it only occurs once, two different suffixes will never have it in the same position), and won't "overflow" into the other string.
Now, it can be seen that you should be able to see why you only need to see consecutive values in the $LCP$ array (the argument is based on contradiction and the fact that the suffixes in $SA$ are in lexicographic order). Keep checking the $LCP$ array for the maximum value such that the two suffixes being compared do not belong to the same original string. If they don't belong to the same original string (one begins in $A$ and the other in $B$), then the largest such value is the length of the largest common substring.
As an example, consider $A = abcabc$ and $B = bc$. Then, $S = abcabc\#bc$. Sorted suffixes are $\{abc\#bc, abcabc\#bc, bc, bc\#bc, bcabc\#bc, c, c\#bc, cabc\#bc\}$.
$\begin{align*}
SA &= [4, 1, 8, 5, 2, 9, 6, 3, 7] \\
LCP &= [-, 3, 0, 2, 2, 0, 1, 1, 0]
\end{align*}$
Now, the greatest value is $LCP[2] = 3$, but it is for $SA[1]$ and $SA[2]$, both of which start in the string $A$. So, we ignore that. On the other hand, $LCP[4] = 2$ is for $SA[3]$ (corresponds to the suffix $bc$ of $B$) and $SA[4]$ (corresponding to suffix $bcabc\#bc$ of $A$). So, this is the longest common substring between the two strings. For getting the actual substring, you take a length $2$ (value of the greatest feasible $LCP$) substring starting from either $SA[3]$ or $SA[4]$, which is $bc$.
