The transistor is like a capacitor. The charge is stored in the channel. As soon as the transistor is switched off, the charge leaves through drain and source. This phenomenon is called charge injection.
Another problem is capacitive coupling. Switching the transistor causes a current across the parasitic capacitances.
These effects can be seen as spikes and additional charge on the capacitor.
Apart from that you are driving your transistor with an ideal source, add some source impedance to make the simulation more realistic.
Another problem with your circuit is that you did not consider the effect of the body diode. I can't see it in your simulation but the problem is there. Most likely because you are sampling quite fast and the voltages across the diode are too small.