Paul,

Based on your comment, I assumed that the pic basic was reading the stop bit as the MSB of the data. My conclusion is that the pic clock was too fast. So I reduced the OSCCAL from its 3460h original value to 3450h and it worked on the first try! Now IŽll play a litle bit more with the OSCCAL to find the midpoint to (hopefully) garantee that the circuit will be stable over temperature changes.

Thank you for your valuable help!