As I am working on the statistical research for my graduation, I encountered two problems;

1) I am measuring two multi-dimensional concepts, and I used a factor analysis to measure the components. One of these concepts turned out to consist of 6 components.
I regrouped all items into these components and calculated the mean of each group. I used these means to do another factor analysis, to measure if the groups represented the concept I wanted to measure. But now the factor analysis showed there were 2 components.

How do I solve this? Does this mean that the first 6 components are 3rd order constructs, and these 2 are 2nd order constructs?
And should I calculate the final mean of the construct out of these two components and use this mean to do my mediator analysis?

2) My second question is about the reliability analysis.
All the Cronbach's Alpha are above .7, which is good. But when I look at "Corrected item total correlation" and "Cronbach's Alpha if deleted" a few of my items will higher the value of my Cronbach's if removed.
For some, the difference is .001, so I don't think I should remove those, but others differ with .009, which seems like I should remove these. (The corrected item total value of these items is below .5).
So should I do this this way or not?

And when this is done I think the factor analysis has to be redone as well?

Sorry for the long questions, but thank you in advance!