Kappa incorrect when changing number of reviewers

[esyriani]

We are 6 reviewers. We imported 100 papers and screened them. Each paper is assigned all 6 reviewers. We got a value of Kappa. Then we resolved all conflicts.
Next, we imported 1816 papers and screened them. Each paper is assigned 2 reviewers. The Kappa is negative before resolving conflicts but we only have 8% of conflicts. See statistics figure at the end.

At the beginning, we had 2 screening phases: P1 and P2 (final). The first 100 papers were screened in P1. Then we deleted P2 and imported all remaining 1816 papers in P1. Perhaps this also had an effect on top of changing the number of reviewers per paper.