The first use is the normal Bitransitive construction with prefer; i.e, prefer has one volitional subject A, who is the deemer, and two objects, the first of which X is deemed preferable to the second Y, which is marked by to, and must appear after X.
The second use is quite different; there are two clauses: the main one and an adverbial clause beginning with rather than. The verb in the main clause is the simple Transitive (not Bitransitive) prefer, which simply takes one object and awards it the preference.
The rather than clause that explains what Y was, since it is adverbial, can be fronted easily, while the bitransitive construction has strict word order.
The third sentence is simply ungrammatical. But is not that kind of conjunction.
The key is that different modalities of "the same" verb -- in this case bitransitive versus simple transitive prefer -- can get to the same point via different routes.
As a footnote, I would mention that the syntax and semantics of comparative and superlative constructions (like Bitransitive prefer X to Y) is considered to be one of the most complex and difficult in linguistics.