Did you use the same NMR tube? And, is the Tube Peak shape is the same in both cases which can be acertained by phase correcting the spectra by 180 degress and comparing the two tube peaks. Does the two samples have the same chemical shift between the sample peak and tube peak? Is there a possibility of frequency dependent/second order phase corerction parameters being different for the two cases? What was the relaxation delay in both cases if multiple scans were used for better Signal to Noise ratio? Can you have a standard reference Silicon compound to verify the relative phases not leaving the comparison only to test sample and NMR tube itself?

Silicon is a nucleus with a negative gyromagentic ratio. This means that decoupled peaks or peaks that experience a nOe can have opposite phase to peaks that do not have nOe. Were the experiments done with decoupling? Did you use the same phase or use automatic phase correction?