bivariate -logLikelihood plot

[dqq0404]

Hi,
After I did the bivariate analysis and got the plot, I found the -logLikelihood plot seemed wrong.


Below are the parameters of the two pictures:
best_vs_min_AIC best_vs_min_BIC best_vs_max_AIC best_vs_max_BIC
0.35 -9.19 0.29 -9.25
0.51 -9.03 7.77 -1.76
It shows that the lowest -logL is inconsistent with the number of shared variants in the first plot.
How to explain it?