NaN values in ln_ged_best_sb

Hello,

I am assuming that ln_ged_best_sb is the correct PRIO Grid outcome variable for the forecasting challenge. However, I am noticing 37% NaN values in this variable, even in cases for which the other variables are defined. Why are there NaN values and how will NaN values to treated in scoring?

Best regards,
Ben