[M4IF Technotes] Question about chrominance MVs in a 4MV Macroblock

[Luc Martel]

Hi all,
here is my understanding of what's in the standard but obviously I
understand it wrong because I'm getting very strange results.
Let's suppose I have these MVs for a macroblock (in half sample accuracy),
coordinates are (X,Y)
There is no transparent macroblock so I'm using table 7-6 for 16th pixel
position.
(15, 20) (12, -11)
(13, 15) (11, -9)
Then I calculate the MV for both chrominance blocks
X :
roundtab16[16] = {0,0,0,1,1,1,1,1,1,1,1,1,1,1,2,2};
temp = (15+12+13+11) // sum of x component in hp accuracy
temp_int = temp >> 8; //  temp_int = 6
temp_rem = temp % 16; // temp_rem = 3
MVx = 2*temp_int + roundtab16[temp_rem] = 2*6 + 1; = 13
this result would be in half sample accuracy in the luminance image.
MVx = 13//2 -> 7
same thing for Y.
What I'm I doing wrong and is there any explanation on this clearer that
what is written in the standard?
Tkx in advance
Luc