Hugo asked for clarification on the previous post on verisimilitude; I'm not sure what aspect he wanted clarified, in particular, but here is a rough expansion. It will have to be rough because I have limited time at the moment.
First, we need to be clear about 'realism'; there is no such thing as realism simpliciter. We are always talking about realism about a particular sort of thing. Insofar as arguments against a given sort of realism don't recognize this, they are put into doubt, because they need to justify being carried over from whatever sort of thing it is for which they are found most convincing. For instance, the fact that particle realist has problem P in the case of particle physics does not in any way entail that the natural-historical realist has the same problem in the case of paleontology. Arguments against realist views must be justified for each particular case; there can be no free rides. (Similar things can be said about opposing positions like anti-realism.)
Second, it is simply unjustified to assume that theory-ladenness as such is a problem for a realist. Why would it be? The realist about x is simply committed (at least minimally) to the view that x is probably approximately true. The fact that x has to be conceptualized a certain way, put into a theoretical context, seen from a certain perspective, etc., in order to be posited in this way does not affect the realist claim itself at all. Particular sorts of theory-ladenness might be a problem for particular sorts of realism; indeed, it might be the case that every sort of realism, or every sort of realism associated with a particular discipline, has some sort of theory-ladenness that puts it into doubt -- but this must be proven, not assumed. It is rarely even argued outside the context of mathematical physics.
Third, insofar as we have already conceptualized history as being about traces of the past we have already granted some form of historical realism. By recognizing an archive as an archive, for instance, it has already been granted that we have a relevant hold on something of what happened, namely, on what happened so as to leave a trace of itself. It doesn't mean we have a straightforward hold, naturally; but this is not what's at issue, since what is at issue is whether the historian has anything in virtue of which historical realism about something becomes possible. And there is, if history is possible at all. The only alternative to at least a very minimal historical realism is skepticism about the past entirely. Once we grant that we can actually have evidences of the past, we have already granted that the historian has something of what she is trying to get at when she is trying to get at what happened; the only alternative is to deny that we can actually have evidences of the past at all.
Fourth, approximation does not require a unique reference point. In order to say that something is approximately true we do not have to determine what is precisely true. In the case of history, all we have to say is that it coheres with a whole set of traces of what happened; it therefore coheres with what happened inasmuch as that left traces, it therefore approximates what happened. It is not at all necessary in order to recognize something as approximately true that we identify precisely what it is that we are approximating, so long as we can connect the approximation to what we are approximating. Simply to identify traces of the past as traces of the past suffices for this; or at least, if it does not, opponents of realists have not done the work to show it is not.
Fifth, when we are dealing with approximately true accounts, it is simply erroneous to treat them as if they were single propositions. Mutually exclusive accounts, in the sense of accounts that could not both be precisely true, can both be approximately true; such mutually exclusive accounts do not contradict each other, although they would contain propositions that do contradict each other. But this is not relevant to the issue of whether the accounts themselves are both approximately true. It is not necessary to be able to rank such accounts according to greater and lesser degree of approximation to the truth in order to identify them as approximately true; such a ranking would always have to presuppose that we have already identified the accounts as approximately true. Likewise, there is no reason to think such a ranking would be a linear ranking rather than a multidimensional ranking; that is, given two mutually exclusive approximately true accounts, there may be no straightforward answer as to which more closely approximates the truth. They may have different strengths and weaknesses, for instance. They may approximate the truth in different ways. It is simply an unwarranted assumption to think of approximation to the truth as "unification of particular ideas into clearer accounts with fewer principles". It may be found that in certain cases this sort of progress in approximation is possible (I think it certainly is); but none of this, again, causes a problem for the original identification of some accounts as approximately true; progress in approximation is downriver from identification of approximation, and thus can't cause a problem for it. And all one needs for realism as such is simply the ability to identify things as probably approximately true.
Sixth, that we impose narratives on the evidences we have is simply irrelevant to the question of realism unless it is shown that the precise way in which we must impose the narrative is such as to make impossible the particular variety of realism being considered. I am open to the possibility of an argument for this in particular cases; but (1) it needs to be argued or it has no force against the realist; (2) it needs to be argued on a case-by-case basis.
In short, the case to discredit every form of historical realism hasn't been made yet, although I think it fair to say that certain forms have been shown to be impossible, and that there are issues that might be problems for at least some other forms.