Ah Ha! I see the problem... I guess I should have tried simulating it myself without giving you the 'standard answer'. OK, LTSpice looks at the current flowing into the compensation pin to detect steady state. This chip does not have a compensation pin, so I guess it gets upset when you try to measure steady state. Attached is what you are looking for using behavioural voltage sources. This actually gives you the efficiency over the whole operating range of the converter. Let me know if you have problems.
Simon