More or less agree with you up to here...
In my opinion, you need to roughly double that number. You haven't accounted for at least one of the following: the tax credit effect, seasonality, or the slow rise of awareness that the product exists.
Perhaps the third one there is the big error you're making. If you'd attempted to assess "sustained US demand" for Model S in March of 2013 using a naive model, you'd have come up with something well under 20,000 a year -- they actually started building inventory while producing at a lower rate than that. Once word of mouth started getting around, actual sustained US demand turned out to be about 25K/year. It was at 28K/year before Model 3 was introduced.
If you're trying to estimate from current numbers, you're still facing the seasonal headwind and the tax credit effects, as well as being before awareness has started to stabilize.
The ratios still work (since Brazil hasn't opened up shipments yet), but your estimate of long-term order rate appears to be nonsense. This is a period of wild swings in orders due to tax credit expirations, and it's early days regarding awareness.
During the 4th quarter, Tesla said that about half the orders they were fulfilling were new orders (not reservations). That "half" was about 30825. My simple tax credit expiration model says that about 6 weeks of demand were pulled forward, so I suppose you should divide by 19 weeks and get 1622 orders/week.
But as noted, we're still on early days regarding awareness; it took three years (2012-2015) for Model S to reach steady state deliveries as awareness increased -- going from about 18K to about 25K per year -- and it's probably going to take a bit longer for Model 3. At the same ratio of sales increase, the stable delivery rate would be about 2238/week -- of the high end variants only. Including the SR+ should more than double it, for about 5000/week.
Now maybe I'm wrong. But I think the correct rough estimates for long-term Model 3 demand are 5000/week US, 10000/week worldwide -- maybe slightly less. You're estimating half as much!
My estimates do seem to be in line with Tesla's corporate plans. And so far, they've actually underestimated the long-term sustained demand (they underestimated it for Roadster, Model S, Model X, and reservations for Model 3 and Semi) -- it would be a first for them to overestimate long-term demand.
Given the geographic split, I think perhaps it's OK that the US factory is getting "stuck" around 5000/week production, which matches US demand long-run. If Europe retaliates against the US with tariffs, it'll make more sense to produce cars destined for Europe in China anyway.
I tried to look at this another way by looking at the ratio of the total number of cars sold in the $80K price bracket to the total number sold in the $50K price bracket... but I couldn't find either number.
-----
P.S. People ordering white interior Model 3s in the US are facing long, long waits. Tesla has exceptionally few variants of Model 3; until you can get *all* of them promptly, I'm pretty sure we aren't close to meeting steady state demand.
P.P.S. Tesla just pulled a major "demand lever". People in a lot of the country were avoiding buying Teslas because of lack of service centers within a reasonable driving distance, which are now opening. This will unlock a lot of the latent demand and push us towards the long-run 5000/week demand level. My SG&A estimates assume some increases due to these new service centers, though they seem to be doing them in a very cost-effective and intelligent manner.