Cointegration test in VAR

[SX26]

Hi,

I'm really stuck in this project so any help would be very much appreciated
I have a couple of questions regarding the unit root testing for VAR of 5 variables.
when I run the engle-granger two-step procedure I find that the residuals are non-stationary, hence I'd reject the cointegration.
However when I estimate the johansson's test, it comes up as two cointegrating equations. Please can you advise if I should ignore the engle-granger test and go with a VEC model instead?

Second, when I'm trying to determine the lag order in levels (all variables are I(1)) LR and FPE give me 5, when Scwhartz gives 1 and AIC gives 12 as selected lag order.

Which one should I be using?

Many thanks,