Are there any perks to doing it like this? Why not just start off with enforcing mixin of 4 at the beginning?
One of the reasons (maybe the only one, even) could be that without RCT, a transaction's size would grow significantly when mixin is increased, because any non-round amount needs to be split into many denominated outputs and each of those is 'mixin-ed'.
Even though the default is 4, I believe that many exhanges and pools - who do a lot of transactions - would prefer using the 2, until RCT is enabled.
With RCT, we get a one-time bump in TX size, but we can increase the mixin more cheaply because there will usually be just 1 or 2 outputs as we don't have to split into denominations anymore.