Regression analyses and mediation with not normally distributed control variable HELP

#1
I am measuring the relationship between an IV and DV and whether two other variables mediated/predicted this relationship. All are normally distributed. I also collected the 'age' of the participants as a control variable, which turned out to show two distinct sample populations (18-25, and 26+) with an overall skew of 1.24.

Analysing data as two distinct groups shows no relationship between IV and DV but significant with DV and potentially predictor variables. Analysing whole data shows significance between IV and DV.

So my questions are:
1. Can I run a regression analyses when there is no significant relationship between IV and DV
2. Can I run a regression analyses controlling for age even though it's not normally distributed (and the residuals are not normally distributed)? I.e. for hierarchical regression put IV and age in first, then predictor variables in second, etc..
3. What is the common course of events when a control variable is not normally distributed? ...do you split the data into two sets or do you analyse it as a whole but control for it later in the analyses?

Very new to statistics so not sure what the 'correct' and legitimate path is to do in this situation.

Thanks in advance!