VAR Model Lag Length

estimated reduced form VAR.jpg I intend on estimating a reduced form VAR but I am struggling to choose a lag length. Initially my variables all contained a unit root but I first differenced them in order to induce

Based on my estimation sample, the FPE, AIC, HQIC picked 4 lags while the SBIC chose 1 lag. I understand that the FPE and AIC tend to overestimate estimate the lag order even in infinite samples while the SBIC and HQIC are generally quite accurate in in selecting the true lag order. In my case, the SBIC predicted 1 lag while the HQIC predicted 5 lags and can I, based on the fact that the other information criteria also chose 5 lags, consider my differenced variables to contain 5 lags?