Schematic? Maybe something in the support circuitry of the Bosch BMA020 pcb?
I see the board has a wider voltage range than the BMA020 chip itself, so something must be going on there...
Also, be sure there are no external
pull-up resistors activated on the I2C bus connected to the DUE. The pull-ups are already included on the DUE, and are aggressively sized (on the small end). Additional current due to improperly-sized external pull-ups will
destroy the DUE's I2C hardware interface. Anecdotally, I believe that I have destroyed DUE's TWI by just activating the TWI, and inadvertently leaving the logic states low on either/both of these pins, using the built-in pull-ups.
Maybe take a look at the link posted here:http://forum.arduino.cc/index.php?topic=146802.msg1133641#msg1133641
FWIW: I have a I2C DUE system that works reliably with 13 addressable gateways, each with I2C subnets beyond. So, the TWI hardware does work. However, I do buffer the network.