I dont think you have any compensation problem.
Most likely the load capacitor is messing up your DC Operation Point simulation. The amplifier is probably taking some time to charge the capacitor and the DC op point is calculated when Vout=0, when some of your transistors are in triode. There are two ways to overcome this problem:
1) set up initial conditions, setting the capacitor voltage to the expected amplifier DC voltage
2) increase the DC op point simulation time