As Mark has said, it's, in theory, a solvable problem (we know how to do it; the known methods are not simple).
However, by tweaking things around a bit, we can make this problem easy.
My solution is based on Pedersen commitments; those are based on a large prime-sized group (where the discrete log problem is difficult) and two group members $g$ and $h$ (which have no known relationship; specifically, no one knows the solution $x$ to $g^x = h$).
A Pedersen commitment to the value $x$ is the value $g^x h^r$, for some random $r$; properties; we can issue the commitment (by publishing the value $g^x h^r$), and then later open the commitment (by publishing the values $x, r$; anyone can verify that those values give the commitment.
Someone looking at $g^x h^r$ cannot determine what $x$ is (in fact, for any possible value of $x$, there's a value $r$ that would give that commitment value)
The issuer is unable to open the commitment two ways; that is, if he issues a commitment $g^x h^r$, he is unable to find a value $r'$ such that $g^{x'} h^{r'}$ evaluates to the same value.
With that background in mind, he is my proposal:
Charlie sends the following values to Alice:
$x_1$ and $x_2$
Signed commitments to those values, that is, signed copies of $g^{x_1} h^{r_1}$ and $g^{x_2}h^{r_2}$
The random values $r_1$ and $r_2$ (because he already gave the values he committed to, giving him these random values is harmless)
His public key
Alice then computes $x = x_1 + x_2$, and generates a zero-knowledge proof that the sum of the two values committed to by $g^{x_1} h^{r_1}$ and $g^{x_2}h^{r_2}$ is $x$. This can be done by generating a proof of knowledge that Alice knows the value $s$ such that $h^s = g^{x_1} h^{r_1} \cdot g^{x_2}h^{r_2} \cdot g^{-x}$; Alice can generate such a proof only if $x_1 + x_2 = x$
Alice then forwards to Bob the value $x$, the two signed commitments, the public key (so Bob can verify the signatures) and the zero-knowledge proof (which Bob can also verify).
This appears to address the end goal (and fairly simplely; there are some details I only vaguely waved at, however a bit of research will turn those up).