Yes, it's correct. This is implied by the following two facts, which together imply the needed equality.
The longest palindromic subsequence is at most as long as the longest common subsequence of the string and its reverse.
The longest palindromic subsequence is at least as long as the longest common subsequence of the string and its reverse.
Fact 1 is easy to prove: every palindromic subsequence of the string is of course a subsequence, and it's a subsequence of the string's reverse because S1 is a subsequence of S2 if and only if reverse(S1) is a subsequence of reverse(S2), and the reverse of a palindromic sequence is itself.
Fact 2 is subtler. We argue that, given an LCS of the string and its reverse, we can derive two palindromic subsequences whose mean length is equal to the LCS. It follows by an averaging argument that one or both is at least as long.
I'll illustrate the construction process with your example. Write down the common subsequence together with the indexes in the string.
A C B A C
1 2 3 4 5
A B C
\ | /
A B C
5 4 3 2 1
C A B C A
We extract A (1, 4); B (3, 3); C (5, 2)
. We can derive one palindrome by taking the prefix where the first number does not exceed the second and mirroring it: 1, 3, 4 -> A B A
. We derive the other in mirror fashion from the suffix where the second number does not exceed the first: 2, 3, 5 -> C B C
.
A B C
1 3 5
.>>\ />>
| |
<</ \<<.
4 3 2
A B C
Observe that each letter of the subsequence is used exactly twice (one going and once coming, except the middle, which is used once in both palindromes), hence our observation about the mean holds.