ARDL Bounds Testing conflicting results

[londonphd]

Hello,

I hope someone from Eviews team will spare their time to look into this.

Attached is a file where I described the procedure I followed to estimate a ARDL model for Bounds testing. My own results obtained following Pesaran et al 2001 show that there is no cointregration among my variables. However, if I use Eviews' Bounds Testing function, the results indicate there is a cointregration among my variables. I chose the final model based on minimum AIC value. I figured out that Eviews too uses AIC based model selection.

I think, I maybe doing it incorrectly at the bounds testing stage where F-stats is calculated. I would really appreciate if you could clarify where I maybe making the wrong turn.
Thank you!