The time shift matches the 48/44.1 ratio. Sounds like you didn't actually resample it, you just told it to play back at 44.1 which would slow it down (that's the header change). You should also notice a pitch change between the two. Don't do any conversions track by track or they might not match up well.
Question - do you know the real duration of the show? What was the UA-5 set at when it was powered on? (changes to sampling rate on the UA-5 only occur after it's been power cycled). If you start your JB3 before the UA-5 is feeding an optical signal, the JB3 will write a 48kHz header on the file no matter what you feed it, even though it's actually running at 44.1. It will still record all the bits but mislabel the header.
When you're working, one way to spot resampling vs. header changes is the first takes processing time, the second is instantaneous.