1. The syndicate sizes listed are maximum sizes, so comparing maximum size syndicates to minimum size market classes is definitely skewing these results.
OK. Checking the revenue per urban family at the maximum population for each market class gets more reasonable values, 6 gp at it's highest. Still the maximum number of criminals per urban per urban family seems high even if half of them are just carousers. The maximum level is high enough to be noticeable to everyone in the city as it's more than the level in Russia when the organized crime was at it's highest (1/37 urban families).
2. The hijinks table by James K is an admirable creation, and it's cool to see my work reverse engineered the way I reversed Gygax etc. In any case, the way I actually worked out the math was to work out for each level of experience which activity was most profitable, where profitability took into account risk of being caught, punishment when caught, etc., and then make sure that activity had odds and payout such that the character of that level made about his monthly wages. It's a three-page spreadsheet and a bit of a mess but the upshot is that the vast majority of all activity is carousing and stealing. At no level is assassination ever the most profitable activity, so it would be comparably rare. I never actually worked out how much specific activity was occurring of each type, though, I just worked out what would be most efficient to occur.
So the easy way to estimate the income is just to assume that all of the carousers carouse and everyone else is off stealing? With this information the maximum activity can be estimated from the total income, but for limits on individual hijinks we have to split them. For some the limit is based on the supply of targets and on some on the demand for the hijinks.
Supply limited hijinks
The absolute maximum for smuggling is the amount of trade in a month. For a class I market, that's 1 200 000 gp (4000 loads = 280 000 st = 400 levels of thieves = 2000 attempts at level 1), which will net 144 000 gp. This would deprive the city of that much in toll and tax revenue (more than 7 gp per urban family, or almost as high as all of the urban revenue). 10% of this is barely noticeable (7 sp per family).
The absolute maximum for stealing is the total income of the city, around 1 million gp per 20 000 families (3333 loads = 233 000 st = 1667 levels = 8335 attempts at level 1), which will net 600 000 gp. 10% of this nets 3 gp/family which sounds about right for the syndicate income.
Treasure-hunting is pretty impossible to estimate. If it includes tomb-robbing in addition to looting hidden caches and old ruins, the total amount of hidden wealth could be estimated from the age of the culture in the area. This would give a maximum of how much loot has been accumulated and could ever be discovered in the area. Another way to think of this is looking at it as a steady state system; there's a constant flow of treasure being buried and lost and a sustainable treasure hunting industry can recover it at the same rate. All of this would be very variable depending on the culture and current situation (what are the traditions for grave goods, how many people bury their savings etc.). Any archeologists on the forum? My guess would be 1 month's income per 20 years (a generation) per urbanite = 5 sp per urban family per month = 10 000 gp per 20 000 urban families = 3 successful treasure hunts = 19 attempts at level 1. The 10% level would be only 5 cp per family per month, but would mean that the rest will accumulate for future looters.
Demand limited hijinks
The maximum amount of assassinations would depend on the political situation and traditions of violence in the culture. Built on the estimates given above, the maximum rate would be 0.025 to 1 per month per 20 000 urban families. That's only 12.5 to 500 gp a month if most targets are normal men (as is most of the population). For 20 000 urban families the maximum level would be only 2.5 cp/family (= 1 successful assassination of a normal man = 10 attempts at level 1). The 10% level would be only one attempt per month. Probably the level of targets is skewed towards higher levels as the more powerful people would have more need for assassins.
For carousing and the maximum demand would be based on the amount of funds available for this information and demand for it. As for the actual demand for all kinds of secret information, it too would be highly variable and depend on the culture and local situation. The information could be thought as a commodity in itself, used by the buyer for some concrete edge (possibly financial). Back-calculating from the assumption that carousers will be the source for this income, the sustainable (10% level) would be 2 gp per 20 000 urban families (40 000 gp = 410 successful carouses = 2735 attempts or 31 successful spyings = 308 attempts at level 1). The theoretical 100% level for information gathering would be 40% of the urban income, but what is the net effect of that information trade I cannot really guess. Some of it would give a definite economic edge to someone, but also a lot of it would be just blackmail or used for some political or strategic gain. Anyway, unless there's no monetary gain from using that info the maximum could be even higher. If there's no measurable gain the income could be considered a tax like the income from stealing.