It always amazes me how some people make the hardest problems for themselves!

Firstly, which end has the problem? We could have multiple defects in both ends - so where do we start? Secondly, rather than trying to send two 12 bit samples, have you tried sending just 1 12bit sample and get that going first? Also, to minimise defects in the sending side, rather than use data from the a/d, hard code in values - that makes it easy to test to see if those values are received correctly. Those values could be 1V or 2V so that it is easy to measure on the DAC output.
Thirdly, use the simulator to check your code.

Just a note: RB8 is a BIT - what happens when you assign it to a CHAR?
Also:

high |= a&0x0F; // high: 0000HHHH
flag = 0;

shouldn't that be:

high = a & 0x0F; // high: 0000HHHH
flag = 0;

Remember the old rule: divide and conquer!

Big problems are made of lots of small problems - fix the small problems one by one and the big problem shall dissappear.