Across a very small dataset, and as
@LinekersLegs points out Gore and HRC won the popular vote. I tend to think that the broader elections dataset is more informative with respect to whether sharing party ID with the current office holder generally confers a small advantage, though you are entirely within your rights to argue that the presidency is it's own animal and disagree. Plenty of scholars take exactly that position on the grounds of the differences in the nature of the job and its visibility, the unique nature of the win condition (the Electoral College) or both.
I try to treat the small dataset problem like this: I'm reluctant to make inferences based on what appear to be clear trends in the small dataset when those inferences go against conclusions drawn from better data, and so I'm reluctant to conclude that Gore and HRC lost due to fatigue when their flaws as candidates are so evident. You could also argue that Nixon was a flawed candidate that was competitive in 1960 solely due to Eisenhower's coattails (and that won the office despite his flaws because Vietnam rendered a Democrat unelectable in 1968), both of which argue against fatigue as the explanation and for other factors being determinative.
Presidential elections are a messy area of study with comparatively little scholarly agreement because we so infrequently get information that allows us to update our priors, which in turn means that a lot of hypotheses remain sustainable over time.
