Looking at the ADA4898 datasheet the 0.1 to 10 Hz noise trace (fig. 45) is worse than I would expect based on the spectral density plot (fig. 15) and much worse than other op amps with similar voltage noise specifications (for example the AD797, OPA211 and LT1028). What is the reason for that? Fig. 15 almost makes it look like there is a 1/f^2 noise characteristic below about 5 Hz, could that explain the difference? Also, it seems like there is something wrong with the schematic of the low-frequency noise test circuit (fig. 46). It shows the ADA4898 being operated at unity gain with an AD743 amplifying the output but the AD743 has higher (by more than a factor of 2) voltage noise than the ADA4898.
You are correct in your observations. The dominant 1/f noise contributor in fig. 46 is the AD743. The 0.1Hz to 10Hz noise performance of the ADA4898 is better than what is shown in fig. 45. The datasheet, specifically, that section is in the process of being updated.