Isobar collisions which were thought with a same background and different magnetic fields provide an opportunity to verify the chiral magnetic effect (CME) in relativistic heavy-ion collisions. However, the first result from the RHIC-STAR isobar experiment did not observed the CME signal, but discovered that the backgrounds are different between 96 44 Ru + 96 44 Ru and 96 40 Zr + 96 40 Zr collisions. We test eighteen cases of Woods-Saxon parameter settings resulted from different nuclear deformation or nuclear structure effects using a mutiphase transport model. We find out that seven cases can reasonably reproduce three reference ratios measured by the STAR experiment. With considering both the halo-type neutron skin structure and CME-like charge separation, we demonstrate that it is difficult for the CME observables (∆δ, ∆δ ratio, ∆γ and ∆γ/v 2 ratio) to distinguish the presence or absence of the CME, if the CME strength is weak in isobar collisions. It is because the final state interactions significantly weaken the initial CME signal, resulting in non-linear sensitivities of the CME observables to the CME strength. Therefore, more sensitive observables are required to search for the possible small CME signal in isobar collisions.