Engle Granger Vs ECM & VECM

[nannigia]

I have a time series model (39 obs only) that Engle Granger (OLS regression with stationary residuals) coefficients is OK however VECM and ECM G to S models are giving me different coefficients and therefore different results.

Can we say Engle Granger approach is still valid? although it suffers from not normally distributed residuals and Serial correlation? i appreciate if you help me.