Dispersion Optimization Discrepancy in a Dogleg

[Demurtas]

Hello everyone,
I'm working on a transport line with a dogleg for a Compton source. I'm having problems with the optimization of the dispersion. The lattice is composed of two sector dipoles with three quadrupoles in between. I used the elegant optimizer to impose the dispersion and its derivative to be equal to 0 (<10^-7) at the end of the line, but I found a discrepancy between the lattice dispersion, which indeed goes to 0, and the dispersion computed on the beam with sddsanalyzebeam, which does not vanish at the dogleg exit. By looking at the behaviour of the emittance I can see that the beam analyzer is right, indeed the emittance doesn't recover its initial value.
I thought this could be due to a chromatic effect, but even setting sigma_dp = 1.3e-9 in &bunched_beam does not fix the problem.
The plot of the horizontal emittance and dispersion along the line and the files to run the code, which is launched with ./Script, are attached.

Does anyone know what the problem is and a possible solution?