The problem comes with some $e5 at $c002 in page 1. In +3 the page 1 is initialized with zeroes, from $c000 and so on. But in +2a mode, the page 1 is full of $e5. There is routine in $c0b1 in page 1 that initializes some adress in some call opcodes, but if byte at $c002 is not 0 then doesn't set this adresses, because uses this byte as a "already init" mark.
It is not an emulation issue, because this bytes are zeroed in the init of the computer (rom 0), and later it is inited with $e5 from rom 2 (+3dos ), by a ldir at $1a87.