The use of PARAFAC for modeling GC × GC-TOFMS peaks is well documented. This success is due to the trilinear structure of these data under ideal, or sufficiently close to ideal, chromatographic conditions. However, using temperature programming to cope with the general elution problem, deviations from trilinearity within a run are more likely to be seen for the following three cases: (1) compounds (i.e., analytes) severely broadened on the first column hence defined by many modulation periods, (2) analytes with a very high retention factor on the second column and likely wrapped around in that dimension, or (3) with fast temperature program rates. This deviation from trilinearity is seen as retention time-shifted peak profiles in subsequent modulation periods (first column fractions). In this report, a relaxed yet powerful version of PARAFAC, known as PARAFAC2 has been applied to handle this shift within the model step by allowing generation of individual peak profiles in subsequent first column fractions. An alternative approach was also studied, utilizing a standard retention time shift correction to restore the data trilinearity structure followed by PARAFAC. These two approaches are compared when identifying and quantifying a known analyte over a large concentration series where a certain shift is simulated in the successive first column fractions. Finally, the methods are applied to real chromatographic data showing severely shifted peak profiles. The pros and cons of the presented approaches are discussed in relation to the model parameters, the signal-to-noise ratio and the degree of shift.