why satisfactory explanation does not require correct prediction

Truely objective factual explanation is logically impossible because from the idea of any matter of fact, considered in itself, nothing validly follows about the existence of any other matter of fact.

But this does not show that truely objective explanation is impossible, it just shows that it is impossible using the model of deductively valid arguement.

From the point of view of deductively valid arguement anything can be followed by anything, and no logically possible combination of facts is more probable than any other no matter how many facts have been examined.

It may seem that every factually 'satisfactory' arguement we use does as a matter of fact involve a law, and that laws are universal statements connecting matters of fact, arrived at by universal experience. And that universal experience entitles us to suppose that where such a law should be repeated it probably will be. So, consequently satisfactory factual explanation requires a satisfactory law, which requires some probability or guarantee of what will happen.

But on the contrary, every realistic factual explanation is a proto law, not because it will have been found to apply universally in experience, but because if there are two situations that are apparently indistinguishable then they should both produce the same result, otherwise there must be some difference between them after all; i.e. if we are assuming realism in the form that the contents of the situation should themselves produce what occurs. This means we work out from particular towards universal experience, and this is why we often have to examine particular experiences so closely. Interpreting such apparent laws as empirticle laws established by universal experience, which then justifies our prediction or expectation of a similar instance places to high a requirement on interpretation of the particular instance in terms of the suppose 'universal' experience, or model of experience. This divorces us from experience and places these experiences under the requirement of language and interpretation.

But if instead we take the realist approach, because we are concentrating on how the contents of a situation could themselves produce what occurs there is no way these contents could guarantee to us that this approach is correct, even if it is correct. They are not these things in themselves and an information bureau. They are not interested in what we may or may not think of them. So concentrating on some guarantee or probability as the ground of what establishes our explanation as satisfactory concentrates on an aspect that couldn't be present as establishing the explanation as satisfactory. However, if we have an apparently objectively satisfactory explanation, this requires that the situation appears self sufficient, and inspite of many experiences where things happen that are apparently not self sufficiently explainable in terms of their contents, such a self sufficient explanation will seem very convincing and so will seem to justify such things as predictions as to what will happen, and sanction opinions about what cannot happen, because how could the situation occur otherwise than by being self sufficient? But his is no guarantee that establishes the explanation as satisfactory, it is a consequence of what being apparently satisfactory establishes.