Principal components analysis

Hi there,

When I use Principal component analysis from TPrincipal with my data I realize that the weight’s distribution of each component for the set of the data is not always centered around zero. I mean, for each picture/data I have a certain weight related to each PCs, and if I do an histogram of this weight for the entire data (an histogram for each PC of course) I have a certain distribution which is, for certain PCs, not centered in zero. This is strange because if I assume that the eigenvectors of the Covariance matrix are all orthogonals then I don’t understand my observation.
Is it normal? Am I missing something?

Thanks for your help,