Why is it not possible to hold on to both horns of a moral dilemma? Not just because a moral dilemma is defined as such, just in terms of this impossibility.
In the section covering moral dilemmas, the SEP article on deontic logic goes over the idea of waiving agglomeration to yield the proper logical image of moral dilemmas. The idea is that from OBA and OB~A, one would derive OBA&~A, i.e. a moral dilemma would require one to actualize a contradiction, "which is impossible." More saliently, it does not seem as if facing a moral dilemma is emotionally the same thing as being directly asked to actualize a contradiction. If two authority figures on a par relative to you issue contrary demands, you can't compose the demands in your mind into one hyperdemand that is not contradictory, but this is not to say that either authority figure has itself issued a contradictory demand. (Of course, some will be motivated to reject authority figures in principle, perhaps for this very kind of reason (that conflicts between authorities could translate into dilemmas), and one might think that there was "preauthorization" involved anyway, viz. besides there being authority figures, there is also a standing rule, issued by no individual authority as such, to obey authority figures; whence we can infer, from this rule and the individual demands made, a contradictory obligation.)
However, again, why is it impossible to do both things required in a moral dilemma (or all three things required in a moral trilemma, or whatever)? Suppose that we split the OB operator into "it is a yet-to-be discharged obligation that..." = YTB and "it was a sufficiently discharged obligation that..." = DOB. We would say, per a moral dilemma, that it could not be true that DOBA&DOB~A: DOB is a factive deontic operator, after all, so DOBA&DOBB → A&B. But it seems, then, that in forming the logical image of a moral dilemma, one is compelled to agglomerate the propositions on which the deontic operators are operating, i.e. to conjoin them. Otherwise, saying, "Separate obligations to do conflicting things are not equivalent to a single contradictory obligation," would be as justifiable as saying, "Separate conflicting sentences are not equivalent to a single contradictory sentence." Even though there are senses/contexts in which we simply "entertain" pairs of contradictory sentences without thinking of them as one sentence, yet if we say that two sentences are true, we do more than merely "entertain" them, and then we would add them together.
Inverted: or else might we say that DOBA&DOBB doesn't conform to agglomeration, either? This seems to open the door (beyond or around paraconsistency, maybe) to saying that one had resolved a moral dilemma not because one had violated the implied impossibility of such resolution, but because the impossibility in question had been itself defined in relation to the supposed impossibility of actualizing a contradiction, which here was not assumed, neither enacted. In other words, somehow, one had done A as well as ~A, and yet the doing of each did not amount to doing A&~A. I suspect that if we go that far in revising SDL, we will put ourselves in the unenviable position of having to explain why we are concerned about DL period, since by wiping out all the normal logical images of negation, conjunction, and so on, we turn our revised SDL into little more than a game of symbols all of whose moves involve reference to arbitrary intuitions (about the possibility of unsolvable moral dilemmas, for example).