That's simply because the dependent variable in the UECM should be Δy instead of y, i.e. D(logKINBU).

You actually said "the summary result is very different with R-squared that is very high" which could lead you to the solution.