I don't think you're familiar with overfitting and related data fallacies, like simpson's paradox, or surviorship bias.
Actually, I am thoroughly familiar with them. This kind of mathematical analysis and it'ts pitfalls are what I do in my "day job." Simpsons' paradox is one of my favorites.
Do you think SAUL's prediction that Craig posted upthread exactly mirrors the plot from Howie for the .2 isopleth because both sources independently arrived at the exact same curve from independent sets of empirical data?
No, they have similar results because they are base on the same theories of bubble formation in supersaturated fluids, as are every model since Haldane. And, on extensive re-analysis of the same dive data since the collection of novel data is expensive and really only the domain of a major Navy, particularly the US Navy.
Can you point to any theory that fits the data better? can you point to meaningful data that they didn't take into account or that contradicts them?
The data is insufficient to derive precise answers. But it is more than adequate to validate the the principles underlying all current algorithms as (so far) better than all others (so far) proposed. It is more than adequate to determine the monotonic relationship between PDCS and Supersaturation. It is more than adequate to determine that the curve is non-linear and steeper at the GF=100 end than the GF=0 end. It is probably good enough to be fairly sure that the first derivative is also positive and monotonically increasing as well, though I would have to look closer to be sure.
Just because a conclusion from the data is not exactly correct doesn't mean it isn't usefully correct. For an example most people will understand: the fact that Newtonian physics and theory of gravity is wrong at relativistic speeds, does not make it meaningless or less useful in normal conditions. Which is why engineers everywhere rarely use Einsteins physics instead (unless they do orbital mechanics or other relatively esoteric work).
Any hypothesis that doesn't fit the data we do have (eg. 4% reduction in risk going from GF100 to GF50) is more wrong than the ones that do. Maybe as we get more date the hypothesis we have now will be superseded, or even invalidated. But until that data is in hand, you can't just pretend that they are meaningless. And most likely the new hypothesis will be refinements rather than wholesale rejections of what we have now.