The plot looks exactly right to me, depending upon what the original signal is. The discontinuities in the phase angle are due to zero-crossings in the magnitude. However, the phase jumps are less than pi (which is what you might expect if the input signal were originally real and had been windowed when applying the FFT). So my guess is that this signal is some form of complex modulation with some discrete frequencies.
Generally viewing the phase tells you a lot less about the signal than its magnitude. It is essential for reconstructing the signal, but much harder to interpret (particularly as it is subject to dramatic changes with a simple shift in the time domain).
You might be better plotting the phase without using the unwrap command In this case, the issue is the unwrap command - its aim is to limit jumps in phase to be less than pi, which is why the final phase is around -4pi.
I'm not sure what your challenge is, but it isn't clear what you mean by reconstructing the samples. You've already got these, so what's the issue? (Are you being asked to reconstruct interpolated samples)?