Puzzled.
I do trust you have resistors in series with each of the LEDs in the bargraph? 470 Ohms each should do.
So "RX/D0" is shorted to "TX/D1" for a loopback test. The only concern there is that since your transmit and receive modules each presume to perform a complete packet, you are relying on the receive buffer and if the sending was buffered, you would enter the receive routine before the complete packet had been sent in which case it may baulk.
As Koepel says,
I would prefer to see a state machine implementation where sending and receiving are performed simultaneously.
And while it is clearly not your own representation, I wonder why "RX/D0" and "TX/D1" are wrongly labelled? 
That is the "new" forum of this year. 