Dramatic output difference between version 2.2.10 and 2.3.0

[sNO0By]

My attempt is to count k-mer in T2T_v1.1 genome, which is supposed to be 3 billion bp in length.
At first, I used jellyfish 2.2.10 to count 2-mers. However, the result (partly) is as follows:
>19688679
AA
>62474176
AC
>7400287
AG
I then realized the numbers sum up(less than 1 billion) is far from either 3 billion or 2x3=6 billion, and I used 2.3.0 to check the situation. Results are as follows:
>297813223
AA
>152313792
AC
>209422175
AG
This time, the numbers sum up is near 3 billion. You can see how different these outputs are. I am confused at it.