Regarding InSAR Atmospheric Correction

[ZCC20000]

During InSAR data processing, I found that the raw unwrapped interferometric results, without atmospheric correction, were actually better than those after atmospheric delay correction. Specifically, the time baseline of the interferometric pair is only 12 days, theoretically resulting in a weak atmospheric delay effect. However, when corrected using GACOS and ERA5 atmospheric products, both the coherence and deformation consistency of the results decreased. Further comparative analysis showed that the tropospheric delay distribution of GACOS and ERA5 was basically consistent with the GNSS observation results, indicating that the external atmospheric data itself has high reliability. However, in actual correction, the corrected interferogram performed worse. What could be the reason for this?