bug in thermodynamics

There is a strange loss of heat conservation when a) mass falling in the environment (advection) meets b) a pipe.  To make this stranger, this only occurs when c) the pipe is running right->left

Setup 1:

Exchange only occurring at the junction circled.  Petrol temp in: 26.85C (default).  Petrol out: ~33.4C (steady-state).  Crude in: 76.85C (default).  Crude out: ~69.9C (steady-state).

This represents a heat gain by the petrol of 10kg/s * 1.76 kDTU/kg/C * (33.4C - 26.85C) = 115kDTU/s.

The heat loss by the crude is 10kg/s * 1.69 kDTU/kg/C * (69.9C - 76.85C) = -117kDTU/s.

This is good enough, and I regard it as conservation of heat.  But change the orientation right:

Now (steady-state): petrol in: 26.85C.  petrol out: 70.4C.  crude in: 76.85C.  crude out: 35.7C.

The fact that this the output temperatures are different from the above is weird enough (though I've seen plenty of similar asymmetries; I think this is because you process ticks from left->right.).  But more concerning is that heat is not being conserved:

petrol gain: 10 * 1.76 * (70.4 - 26.85) = 766.5

crude loss: 10 * 1.69 * (35.7 - 76.85) = -695.4

This discrepancy is much too large to attribute to rounding error.  (Indeed, I first encountered it in a much larger context of trying to instrument / optimize a heat exchanger.  It took awhile to isolate the source of the discrepancy.)