I believe there is no physics reason why U can't be depleted to decrease entropy. There is an assumption in Entropy/Life discussions that life is a cyclic process, thereby having the Carnot cycle efficiency losses. But if there is a life form who's "cycle" always increases the number of copies with high efficiency, then the net effect might possibly be like it is staying in the "upper half" of a Carnot cycle, continually decreasing entropy of the Universe as it reproduces, at the expense of potential and kinetic energies. The return path ("bottom half") of a Carnot cycle is where it would normally become apparent that more entropy was generated than lost. Using energy to decrease entropy is a common core concept, so I do not know why I can't find a book on exactly what I'm saying, i.e. model replicators as continually expanding rather than thinking only in terms of recycling matter on Earth. If you assume the number of copies has an upper limit, then yes, you're going to conclude that it's cyclic and that net entropy is being created despite the local negative entropy in individual organism.
I've always felt that if anything had objective "moral rightness" to it, then it would be "efficient use of energy". There is also an instinct in everyone that says "disorder is bad". I could rationalize that this is because people need it for things, so it's a natural human instinct, and a valid viewpoint if survival of the species is the highest good. But I have no thought of myself or humanity in having this intuition, and selfishness for the species (assuming the benefit of the species is the highest good, humanism) is no more valid than selfishness for oneself which is no more valid (in terms of "moral rightness") than a selfish gene.
OK, so combining the thought and assumption that we should worship (i.e., assign a moral rightness to it) the thing that made us who we are (in terms of both body and intellect ... evolution of genes and memes) and assuming that the two intuitions above it has endowed at least me with, I come to the conclusion that using the most energy to move the most matter to make the most survivable copies, whatever those copies are, is the highest good. Since only a few of my ancestors were human, maybe I should not have any qualms about my children not being human, if they are the most powerful. But "powerful" is not it. Efficient is the goal. So it's not the most energy nor the most matter, but it is the most copies. In fact, it is to get the most copies with the least amount of energy and matter, but letting the most efficient type of copy be the one who gets the most energy and matter. So, in looking at the result of what this morality might be, I come to the preliminary conclusion via my first paragraph that the ideally efficient replicator can reduce the entropy of the Universe at the expense of stored energy potentials. Lowering the temperature may also be in this morality, as it brings things to a stop which will keep the copies in their final state. This does not automatically mean the "heat death of Universe" perspective is wrong, as "random" is often a matter of perspective, nor that a Big Crunch would not occur.
I can't say anything more explicit or clear, unless I do more work to give an exact example such as starting with a 50% efficient solar cell connected to a 90% efficient motor and start moving atoms around and keeping a measure of system-wide Gibb's free energy, internal energy, and entropy.