i’m fairly certain way the arrows work is by design and is not a bug.
if i’m remembering correctly, this would have been implemented at the same time the observation page was changed to show only month and year for the dates on obscured observations. this was all done to make it harder to pin down the time when obscured observations were made as a means to prevent interpolation of coordinates.
if there is a bug or inconsistency in all this, it is that in your first screenshot above, the two obscured observations in the middle show “2d” and “5d”. that would pin the observation date for down to a single day or two rather than a month.