take a look at this Octave documentation:
http://www.gnu.org/software/octave/doc/interpreter/Linear-Least-Squares.html
In the description of output beta, it says that the value will be the pseudo-inverse of x times y when the matrix is not of full-rank (as is your case for matrix [1, 1]. [0.5; 0.5] is the pseudo-inverse of [1, 1].
Hope that helps!