The Elmegreens examined 269 spirals in the Hubble Ultra Deep Field and discarded all but 41 because of factors such as an inability to discern a clear spiral structure or the lack of redshift data which gives a galaxyâ(TM)s age.
They divided these 41 spiral galaxies into five different types, based on features such as the number and clarity of arms, whether well-defined or clumpy and so on.
It sounds like they only found a few of each type, seems more like a good hypothesis than "the answer". It also makes you wonder if they cherry picked some of their data.
Imagine that you're attempting to determine when spiral structure typically arose.
1. You throw out all non-spirals: not relevant.
2. You throw out proto-spirals where there's mushy arm-sh structures: potential bias, yes but
2a. You also throw out other spirals where you cannot objectively classify them as grand (2) or multi-armed (>2) spirals or... to one of the five types -- not an inherent time bias.
3. You throw out all data where you have no redshift to determine age: potential bias, yes but
3a You're attempting to determine a relationship with age. If you have no age data, how is that cherry picking?
There is a difference between objectively screening data based on logical considerations and cherry picking. Cherry picking typically involves biased selections or the use of supposedly objective selection criteria to obtain a directed result. I say supposedly because the true objectivity depends upon how the selection criteria actually relate to the hypothesis or analytical method.
As for the rest, I don't see how the paper claims to have "the answer." You're also incorrect that it's a good hypothesis -- the hypothesis is what you test against the data, not the conclusion that your observations are consistent with the hypothesis. They have a decent conclusion of consistency. Now they could use independent confimation, hopefully with a larger population of samples.