>I don't know how much an ion engine costs. I imagine it isn't very much compared to the mission cost.

This is actually not correct: SEP is incredibly expensive relative to chemical counterparts & tends to have a higher operational cost; & the difference in cost can be significant even in Discovery-class budgets (its actually crippling in smaller budgets). The increased costs appear on several budget entries:

First of all, the drives themselves are expensive to build & inherently heavy relative to chemical thrusters, so you lose a significant part of the mass advantage that you gain from the less fuel you have to carry;
Secondly, Xe fuel is >bloody< expensive, to the point where using SEP for any kind of low budget mission - we were offered a free SEP to test drive for a simple asteroid mission but found that the cost of fueling the thing ate up a big slab of our budget - as a primary thruster is prohibitive, although bolt on guidance SEPs to replace hydrazine thrusters are still feasible;
Thirdly - as you yourself identify - the power requirements of SEPs really call for solar panels on the edge of state of the art & that level of technology is never cheap... it also tends to raise deployment issues which in & of themselves tend to turn mission designers' hair white;
Fourthly, being low thrust, SEP-driven craft require an old fashioned level of ops & nav to keep on course, which chemically propulsed craft can often sidestep with hibernation or beacon cruise software... a significant increase in cost with limited time available on the DSN; &
Five, if you're trying to actually orbit something, you'll almost certainly need a chemical engine anyway for the OIM... SEPs can enter orbits (see SMART-1); but being a low thrust drive, the process is almost as long & tedious as it is for solar sails to do the same thing.

As a general rule, the more complicated - & therefore: more expensive - a mission concept is, the less pricy relatively a SEP component becomes; & the reverse is very definitely true. That's why i keep pointing out that - while SEP can certainly make difficult missions less difficult, it doesn't to help us fly simple ones; so its uses are fairly limited in a political environment when budgets are definitely tight.

It's worth noting that the SEP-driven Dawn owes its conceptual heritage >at least< to Diana, a SEP-powered Moon/Wilson-Harrington mission that Dr Russell put up in the first Discovery assessment (the one that first Lunar Prospector & later Stardust won). NASA certainly could have been more proactive in embracing this technology; but the challenge of selecting between a very good mission without the risk of SEP & a possibly better one with that risk is undoubtably a challenge.

Robert