I wrote a blog post on prediction markets, and specifically some of the problems with a popular conception of prediction markets that I've seen out in the wild. It may be of interest to people on LW, so I am including a link to it... here. (Robin Hanson already commented, and he didn't seem to hate it, so I feel pretty good about it already)
I think your point is correct that if there is only economic exposure on one side of a market, then it affects the interpretability of the market prices, as it then becomes an insurance market which requires a premium for the other side of the trade. (With normal insurance, you pay the premium upfront and the insurance underwriter invests that money for earnings, so insurance prices are actually much closer to the actuarially correct price than one would naively expect.) Depending on the size of the market, though, the premium could be small.
I agree that a market COULD be formed without symmetric event risk, I just think it's unlikely that we will see one formed. I think the symmetry makes a market much more likely, and economics is first a social science, so proving that something is possible is far from proving that it will occur. A market has costs to operate, so it has to have a compelling reason to exist, and bringing together natural participants is one of these core reasons. Another factor that would make a public market more likely is a larger number of smaller participants (there were a large number of small farmers when the commodity markets were established, for example).
Probability of a public prediction market forming is increasing with: number of participants exposed to an identical event, balanced natural exposure to the event from each side, and accuracy of the forecasts for the impact on economic outcomes of that event (if your price goes up, your profit goes up in an easily forecastable way). Most futures markets have all three of these, but since events that are hard to forecast (and thus need prediction markets) also have impacts that are hard to forecast, prediction markets so far seem to have hard time scoring high in that third category. Most binary events that have economic impacts are also either broadly good or bad, which makes the second category difficult. Part of this may be lack of imagination on my part, of course, and as I said previously, I would like to be proved wrong.