there is a certain amount of short-term churn: people who signup and then leave again. if that rate hasn't changed we can ignore them, since the number leaving this month is made up by the number arriving now who will leave next month. it comes out in the wash.
then there's the slower loss of long term users, which needs to be balanced by keeping a few new signups.
so far, the above is more-or-less consistent with what you say. but then you end with a comment that somehow mixes the two, trying to explain a mismatch in the second process with people who "even out" in the first.
it would be better to say that one or both of the following (which describe decline in terms of the two processes outlined above) is happening:
A - the amount of churn is falling off. so the people present for only this month are less than the people who were only there last month.
B - long term uses that drift away aren't being replaced by new users.
because (B) eats into the entire user base, while (A) is relative to the new users per month, only (B) can explain a sustained decline (the first process can't explain a loss of more people than sign up per month).
so we can simply wait and see. if the decline is prolonged then it must be long-term users that are leaving and not being replaced.