I run both mlogit and mprobit regressions. Stata reports pseudo R2 for mlogit but not for mprobit.

I understand that McFadden R2 is calculated using 1 - loglik(intercept only model)/loglik(full model). So I try to calculate R2 for mprobit model using the same method. The problem is:

When I run "mlogit depvar indepvar, base(0)", using fitstat, Log-Lik Intercept Only: -926.720
However, when I run "mlogit depvar, base(0)" this supposedly intercept-only model gives log-likelihood of -1199.8673.

If I manually run intercept-only model and calculate the McFadden R2 on my own, I couldn't get the same Pseudo R2 as shown by Stata. This raises my concern of calculating Pseudo R2 for mprobit model on my own since the figures won't be comparable.

Is there anything I did wrong? Or anyway to obtain Pseudo R2 for mprobit?