Zhenhao is right. Most likely the problem is that your T1 scans are to different from the T1 template to allow segmentation to work correctly. Manually reorienting to the AC is one option to correct for this. Additionally you could use some optimized coregistration method. Here you would first coregister your subject's T1 scan to the template T1, and next you subject's (mean) functional scan to the template T1 (make sure you keep all other functional scans in allignement with the functional scan you are reorienting. Than coregister your subjects T1 scan to the (mean) subject's functional scan(s). Than you can do the segmentation and apply the normalization parameters to the functional scans.
Hope this helps.