i think the third up to the last questions all depend on your theoretical framework. even without telling me, what does your TF say about predicted variable relations? as to the first question, descriptive stats help you analyze normality and hetero/homogeneity of your data. as to number 2, you need to report important stats and provide appropriate interpretation
AMOS is very limited, you should swap to lavaan package of the R software (is free), this includes WLSM, WLSMV and DWLS estimators that dont have the sample issues that WLS (ADF).