Dixon Magnetic Resonance Fingerprinting (Dixon-MRF) has been emerging for achieving efficient fat suppression in body applications of MRF. Fat quantification using traditional Dixon imaging relies on a pre-calibrated proton density fat spectrum. The fat spectrum is known to be composed of different fat peaks, which have different relaxation times and are affected by J-coupling modulations. Existing Dixon-MRF methods perform water-fat separation assuming a constant pre-calibrated fat spectrum and not accounting how the complex fat spectrum varies along the MRF flip angle train. The present work proposes a framework to characterize the effect of fat spectrum complexity in Dixon-MRF.