Physical Review resubmission process imitates Dragonlance

I am currently trying to resubmit one of my articles to Physical Review A. My collaborators and I changed several of the figures (we needed to increase the font size of the axis labels) so I need to upload all the figures and indicate what number figure each of the new ones is. I keep writing down what the correct number and subpanel is (for example, 8b) for each .eps file, but when I go to the next screen in the submission process after uploading, most of them are listed as figure "2" and I am getting an error indicating that I have multiple figures labeled as "2". (This has happened on multiple iterations, and I swear there are a couple of figures that I've corrected more than, um, two times.) WTF? When did gully dwarves start working for APS?