Sadly I cannot make much of a comment on matters of biology, and lately it would seem I rather poorly understand the philosophy of science therein. Nevertheless, and with no promises of effectiveness, here is a point I noticed in the preamble to the question at hand:
OP is trying to argue which, between evolution and design, “better fits” the data. There is a number of ways to interpret this phrase, and with maximal charity many could just assume that it means exactly the criterion one might oneself find paramount in judging how scientific some given idea is. At the risk of becoming uncharitable in so doing I shall not make this assumption, and instead present something that more precisely matches my own sensibilities on such issues, whether they ultimately align with OP’s well or poorly.
Surely, in the most naive sense, “fitting the data” is trivial to achieve better than any scientific description ever could. Suppose, if you will, a god who has full control of every aspect of nature and makes it at every instant behave exactly the way it does. A scientific description may approximate some mechanism arbitrarily closely, but it is trivially easy to just say that god makes it this way, and for no reason we are privy to above and beyond just wanting it so. That our theories map to the data is on this view a reflection of us attempting to model the data in a logically consistent way, but ultimately not of any underlying nature of things, for that is entirely due to god’s whims, which happen to be to make things exactly as they are. An account like that stands in no conflict with any data, of course, and it never can. It is, from the ground up, designed to be unfalsifiable. If it turns out that some measurement the outcome of which we explained as being god’s will is some other value after all, we could either say that god changed the universe, or rendered the measuring results what ever they were in all cases, or makes us conclude things exactly as we do. It certainly accounts for such errors, and in a colloquial sense one may even call them “explained” in this manner. There can be no discrepancy between any data and a model as incorrigible as this. With regards to fitting the data, every scientific model is always going to remain inferior to “god makes us record the data exactly as we do at all times”.
Because such a naive criterion makes devising the perfect model trivial, a case can be made that it is no healthy metric to evaluate the merits of all other models. After all, if a perfect one is so easy to propose, why would we ever waste any time at all to considering anything else? We must therefore either introduce some nuance into what we mean by “fitting the data”, or identify a different criterion that lets us meaningfully grade falsifiable models of nature. In my opinion, whether we consider it nuance or a different metric, the key is predictive power.
Our vulnerability and, ultimately, mortality, is the prime motivator of all our instincts. That’s not to say all other feelings are illusory, but rather that, if we follow the chain of motivations for any given sensibility, we end up with some form of well-being, which in one way or another maps to a well-being of the physiological sort. We are the heirs of beings who succeeded in the struggle for survival, those who evolved to function despite all natural threats to their continuation. Science, being a thing we do, is ultimately one of many means by which we make our lives more stable. In order to outmaneuver threats we must understand their maneuvres in the first place. Understanding how nature works means, under this admittedly crass reduction, knowing what to do in order to survive in it, be it personally or collectively. A given about our condition is that we have memories of the past but none of the future. So in order to make decisions that further our well-being we must find a way to see the future in some way with just the tools at our avail.
This is science. The best scientific description of a phenomenon is not so much one that perfectly fits all previously recorded data – that much, as discussed earlier, can be achieved easily by simply over-fitting – but rather one that consistently ends up fitting data it was not constructed to accomodate. An unfalsifiable model is bad not because it is a poor fit on the data, but because one cannot logically derive dependable predictions from it. It is bad, because it is useless. If we cannot employ it to foresee the future, then it cannot service the advancement of our ultimate goals of mastering the future’s threats and opportunities.
This failure is easy enough to see with the exaggerated manipulator god idea I presented earlier: “What ever observation we shall make in this experiment is going to reflect God’s will for it, just like everything in the past already has” does not tell us what to expect specifically, so we cannot devise strategies to prepare for it. But the common designer account of similarities between various living organisms also falls short of meeting this criterion, for it is an account of all biological features, and as such cannot discriminate between expected and unexpected ones. To account for any specific ones one must begin modeling minutia of the designer’s will.
But at that point one might as well dismiss the designer and model the minutia of natural processes directly, for the bundling of such things into a person-like entity adds nothing to the model’s fitness to predict experimental data. If anything, it over-complicates the model, providing thereby more opportunities to err.