Ah ha. Well that's what I suspected might be necessary, if the deobfuscator would make sure naming was consistent across class files if you process them in batch. Do you know if this is the case, and did you have any success with recompilation using your method? It might be worthwhile for me to add batch jar file processing, as decompressing and then recompressing class files seem to be pretty inconvenient, as you alluded to.
Actually, the first thing I did was run the jar files through a disassembler, but my goal is to be able to add new features into the original Kindle software, and attempting this in Java VM assembly isn't my idea of fun.