Can anyone help with why fmlogit does not accept aweights?

I am running a very basic regression on composite data using fmlogit. At this stage I am only interested in overall proportions and not any differences due to an independent variable. I am trying to use Stata's inbuilt weighting scheme aweight but it does not seem to work with fmlogit. This is my code

fmlogit a b c d [aweight = n], difficult iter(20)

In this data set a + b + c + d = 1 and n is sample size

With aweight I am trying to reflect that the observations (each row) in the data were based on varying sample sizes and so trying to give more weight to the larger samples.

I could use my own weighting scheme where each weight = n/(sum(of all n)) and then use pweight which seems to be accepted by fmlogit.
However, I am not sure if internally my proposed scheme for pweights may not be statistically legitimate.

Thanks for taking the time to read this.

Don