Paired t-test vs one-sample t-test

Hello Community,
I tried to run a paired t-test by using the connectomestats (I have one group and two time-points).
I used the same design matrix of Randomise (Two-Sample Paired T-test (Paired Two-Group Difference) on Randomise/UserGuide - FslWiki)
After, I have also tried to run the same analysis by calculating the difference for each matrix (e.g. for subject1 (baseline-point2) etc…) and run the one-simple t-test as described in FEAT: FEAT/UserGuide - FslWiki

I remember that when I used Randsomise, I had the same results for both analysis, but by connectomestats I have very different results.
Am I making any error?