homoscedasticity violated in more ways than one?

Hello I need some advice, I am using multiple regressions in SPSS. My data is not entirely normal in its distribution but not bad enough to make me wince with fear. All my variables are measured upon a 1-7 likert scale. However upon running my regressions I am getting two problems, one of which may be my own doing, so Ill start with that one.

I have been told I need to control for a variable, because it is theoretically bipolar to one of my predictors. So where in the model should I enter it? can I use stepwise still or do I use the 'enter' function to force it through?

Secondly is the residual scatter plot is strange must be violating but unsure of how to interpret or correct it.

Thank you