Part of their mistake was to underestimate how fast cellular would develop, which suprised just about everyone, but an equally big part was to disregard what pricing was due to technology and what to politics.
Actually, long distance rates plummeted more due to regulatory changes and fiber optics than to competition. For the past century long distance had been deliberately overpriced to subsidize local service and (in places with PTTs) other bits of government bureaucracy. The mistake there was not to realize that with a stroke of a pen those subsidies could be and were removed, which is the main reason that a call from the US to the UK or Hong Kong now costs 2 cpm rather than a dollar.
True, but unlike Iridium vs. cellular, satellite vs. internet radio is not an apples to apples comparison. With telephony, the question is how you get a 3 KHz low-latency full duplex channel (not exactly, but close enough) from one point to another. Satellite really broadcasts, but internet radio fakes it with a separate connection to each recipient. (There is real Internet multicasting but it's a pain to set up and is only used in the geek community to broadcast IETF meetings and the like.) With broad, the question is how you get the same one-way signal to lots of recipients.
This means that it's a question of scale. With the current low numbers of listeners, Internet has the edge as you note due to its parasitic carriage.
Two-way radio spectrum is far from free. 3G definitely works, WiFi is OK for short distances, WiMax is grossly oversold for other than fixed point to point service. They're swell for telephone and individual data service but they're way too expensive for broadcast. Back around the turn of the century, there was what you might call telephone radio, with concerts and the like sent over phone wires to large numbers of listeners. (It was really popular in Hungary for some reason.) As radio developed, radio blew it away because there was no incremental cost per listener, and the phone wires could be used more profitably for telephony. If you use any Internet technology for radio, you're in the same situation, using point-to-point bandwidth for simulated broadcast.
If the total number of listeners to your station is small, in the thousands, point-to-point looks good because of the low cost of entry. But if satellite radio does what its backers hope, and has millions of listeners per station, which is not implausible considering how many listen to Howard Stern on normal broadcast, satellite wins big.
I think the real outcome will depend on questions like whether the satellite radio stations are able to bribe car makers to install receivers as standard equipment in cars so users need only call up and subscribe, no installation or visible startup cost involved. It'd be like cell phones are now, using the equipment as a loss leader made up from subscription revenue. It looks to me like the incremental cost of a Sirius or XM receiver and antenna would be about $100 which is well within the range that cell plans subsidize.
R's,
John