Probably not a glitch, just an indication of the large scatter in the available data due to different testing conditions. E.g., sample geometry, notched vs. unnotched, electropolished finish vs. mechanical grit, alloy variation, stress frequency, stress ratio, etc. That is why S/N curves are often shown in books as broad bands.
MIL-HDBK-5J, Re Al 7075-T6:
“Figures 3.7.6.1.8(a) through (h) provide room-temperature fatigue curves for T6 temper products.”
Eight figures are given, corresponding to various sets of sample geometry and finish, and each figure has 4 or more best-fit curves for different stress ratios or mean stresses. Standard deviations are about 1 log unit (an order of magnitude) even for data obtained under a set of standard conditions.
Hope this helps,
Ken