Negative item-total correlation warning message

[cogcardinal]

Hello,
I have a dataset with 30 dichotomous items. When I run the reliability analysis, I get a Note that says two of my variables "correlate negatively with the total scale and probably should be reversed." When I look at the item-rest correlations, these particular items are positively correlated with the total. They are the smallest correlations in the set, but they are still positive. If I do reverse code these items, the item-rest correlations become negative but the Note goes away.
I'm wondering how these Notes are calculated. I thought perhaps they were based on uncorrected item-total correlations, so I looked the correlation between the problem items and the sum scores. These correlations were positive as well. I'm encountering the same thing with other datasets. The Personality Questionnaire from the lsj-data library is also giving me inconsistent reports between the Note and item-rest correlations.
Thanks in advance for any feedback!
Edit: added screenshot of before/after reverse scoring