That's a fair question, and I certainly don't have all the answers, just making the best choices I can on the balance of probabilities as I see them. I welcome discussion on this so by putting our heads together, we can continually improve the risk reduction.
Also, most steps to try and reduce immunogenicity are less than perfect, and may impart some additional risk while also reducing it. The question is whether it helps more than hinders.
The Liquid/Air interface is a factor in generating aggregates, which are a primary cause of immunogenicity.
I see it like this.
When in the vial, the entire large surface area of the top layer of the liquid is exposed to air.
While in a filled syringe, with an attached liquid filled filter+needle, there will be continuous fluid present all the way to the tip of the needle, only exposing that very small area where air meets liquid inside the needle shaft. It's really the same thing as is happening in a multi-dose injection pen.
And yes backfilling is certainly less than ideal, but as I pointed out before, done quickly with some common sense, it's no worse than filling a syringe from an ampule with its top cracked off, or even having to connect a luer lock syringe and needle out in the open exposed to unsterile air.