Only step 3 is important.  The IV that is signif. there is the one that is important for interpretation. For a Ref. see Kutner et al Applied Linear Statistical Models 5th edition McGraw Hill Irwin.  Best wishes, David

Rany -

You don't want to use the interaction term alone. 

You can forget about an artificial "significance" level.  Also there can be other interactions besides collinearity which obscure the situation.

If each independent variable individually makes sense as a predictor from a subject matter point of view, then the whole model might make more sense.   You can check performance of any model, but results are sometimes spurious based on a given set of data. 

You can graphically compare predictions for different model choices, using residual analyses, but if you do not save out test data, you might overfit.  You might want to research the terms "model selection," and "model validation."  Also you might look into the difference in the use of regression for either explanation or prediction, to see which interests you most.  There is a good example of that found at the end, in the appendix to the following:

Galit Shmueli,

 https://www.researchgate.net/publication/48178170_To_Explain_or_to_Predict

(One confusion that can occur is that the estimated variance of a prediction error can be inflated by bias, the way it is estimated.) 

David mentioned Kutner, etal., 5th ed.  That must be the latest edition of Kutner, M., Nachtsheim, C., and Neter, J.(2004), Applied Linear Regression Models, 4th

ed., McGraw-Hill/Irwin, which I've used before, and agree it's a good one.  There are other texts which might help if you can find a regression or maybe an econometrics textbook, and just check the index and table of contents.  There, and on the internet, you never know what you might find.   :-)   

Cheers - Jim 

Article To Explain or to Predict?

 Thank you so much David Eugene Booth and James R Knaub for your help and for the enlightening responses! I will continue as you both suggested. Thank you for the valuable references as well.

all the best, Rany