What does the grand mean scaling in design section do in two sample t-test in VBM?

Dear Colleagues,

                      I hope this e-mail finds you well. I am doing a VBM analysis. I used a two-sample t-test and TIV as a nuisance factor and when I use the grand mean scaling in the design section, the results are significant. However, when I do not use it, there are no significant results in the same contrast. 

he yellow highlighted one is the one that when I choose, it gives me significant results in contrast (1 -1) but when I select no, it becomes non-significant.

How is it different from this one? When I use this one, it gives significant results in the opposite contrast (-1 1) of only two voxels!!!

My Question is

1- why does this happen? Can I proceed using it? When should I use it?

Are there other resources to understand this matter as I could not find one that explains scaling by subject???

Thank you in advance