Why do you expect the cointegrating coefficient to be 1 or -1?
The reason that you cannot apply JOHANSEN’S method to your time series is because the matrix [tt:3pmh7wka]s11[/tt:3pmh7wka] is singular and cannot be inverted.
I got the same result in R.
> ca.jo(b, ecdet = “none”, type=”eigen”, K=2)
Error in solve.default(S00) :
system is computationally singular: reciprocal condition number = 1.61279e-16