multiple mediators and multiple dvs - different results AMOS and MEDIATE (Hayes)

#1
I have a path model (longitudinal data): 1 independent variable (iv), 4 potential mediators (ms), 4 dependent variables (dvs); I compared models (chi square difference test) with and without direct effects of the iv on the dvs and the models did not differ, i.e. I chose the more parsimonious model without the direct effects.
Some mediation paths were significant in the AMOS model; I checked these with the macro MEDIATE by A.F. Hayes (entering the remaining three dvs as covariates). The estimates are COMPLETELY different from the ones in my AMOS output. Why is that? Is it because AMOS correlates the residuals and MEDIATE uses the variables themselves? I checked the results from MEDIATE also without any covariates; the results got closer to my AMOS results, but are still not the same. Also; What do I make of a path that is significant in my AMOS model but insignificant according to MEDIATE?
Thanks in advance!