The replys probably are echos, the data is transmitted throug one wire...
You are correct, you will get echo. My answer for this was to read the bytes available until they equal the amount sent. In the first part of the code, I sent this:
byte message[6] = {0x81,0x12,0xF1,0x81,0x05};
To get around the echo problem, I read 5 bytes, if the fifth by equals 0x05, then the next byte is coming from the ECU.
The protocol has a tolerance of +-1ms.... Is there a way to test?
I started with "extra" time between bytes sent and I gradually reduced the delay between bytes and I watched for loss of communication. I used a potentiometer in my sketch to make my timing adjustable on the fly.