Case 1: Anyone in the grips of addiction will have preferences whose satisfaction (another shot of heroin, another whisky, another hand of blackjack, etc.) will make them worse off.The second case is incompetent and stupid, an obvious straw man. Lacking relevant information (or having incorrect information) about objective reality can make the satisfaction of any and every preference result in being worse off. If I prefer to cure cancer, and I believe that killing all left-handed people will cure cancer, does that mean that actually satisfying the preference to cure cancer is itself bad? Hardly.
Case 2: Anyone lacking relevant information will have preferences whose satisfaction will make them worse off.
In addition to thus being a vacuous argument against preference-satisfaction (but a good argument for having correct information about the world), one can find that the agent is worse off only by consulting that agent's preferences, those that go unsatisfied.
(It's also egregiously stupid to look at any ethical or meta-ethical preference-fulfillment philosophy as about the fulfillment of individual preferences in isolation. It's trivially obvious that people have competing and mutually-incompatible preferences; preference-fulfillment philosophers simply observe that people have meta-preferences, preferences about other preferences, which they use to decide between competing preferences.)
The first case (addiction) is a little better. Being an addict myself (tobacco), I have some first-hand information about the situation.
It's important to understand that the trade-off between short- and long-term preferences is actually a trade-off between the preferences of two different people: present-me and future-me. When I smoke a cigarette today, I'm fulfilling my preference today to avoid the suffering of withdrawal symptoms at the expense of my "altruistic" preferences towards future-me, a different person. Present-me — the me that decides to smoke a cigarette — is better off in every sense of the word: I presently value the avoidance of withdrawal symptoms more than my altruistic feelings towards future-me.
There's no objective case to be made regarding one's obligations to one's future self. One's obligations to one's future self are completely determined by one's present empathy for one's future self, a different person. We can conclude that one is "worse off" by satisfying a present preference at the expense of a future preference only by arbitrarily privileging the future preference. But if we arbitrarily privilege the future preference, why should we not sacrifice everything, all our lives, just in order to maximize the benefit on the last day — or even hour, minute or second — of our lives? If we are not to absolutely privilege our future preferences, then how much we should privilege them becomes exactly what meta-ethical preference-fulfillment says: how strongly do we feel today about our future preferences?
[h/t to faithlessgod]