I’ve been involved with professional image scanning and editing for reproduction since 1990. There isn’t an effective way of doing this yourself, without building a state-of-the-art photo lab. As you’ve discovered, high-quality scanners are almost all drum mechanisms, and there ain’t no way to wrap a glass neg around a drum. I understand your reasons for wanting to do it yourself, but your desire for ultra-high-quality reproduction without involving an outside service provider is fundamentally incompatible with working from glass negs, particularly at the size you’re talking about.
There’s a couple of problems. One is that flatbed scanners almost all use charge-coupled device (CCD) scanning elements, which do not typically have the dynamic range of the photo-multiplier tubes (PMTs) used in high-end prepress scanners. Another is that, in order to scan a transparent original, you obviously have to have some sort of light source behind the transparency. The low-end desktop scanners like the EPSON you referenced use a translucent sheet (basically a lightbox in the lid over the scan bed) that provides a stationary, diffuse light source. This makes for a much simpler and cheaper mechanism than the point-source illumination units used in higher-end flatbeds with transparency options, like the Imacon *Kamandi pointed out; in those scanners, the illumination provided is much brighter, providing for sharper image capture with additional dynamic range, but the mechanism for moving the illumination source precisely in synch with the scan head is much more expensive. Another problem is that many of the best flatbed transparency scanners use mechanisms for feeding the transparency into the scanner that may not work well with glass negs (I’m thinking of the magnetic frames used in the Imacon units, which have to be sized to the dimensions of the transparency, or the interchangeable trays in the Agfa DuoScan units and others of similar design).
Having said all that, there’s a couple of options open to you. One would be to go ahead and buy a scanner like the Epson above or a used Umax PowerLook III, where you simply slap the tranny down on the glass of the scanner, and see how well you like the results. They may be good enough to satisfy you. They probably wouldn’t pass muster with the photo editors at National Geographic, but ultimately you have to decide what the crossover point on the cost/quality curve is for you.
The other option, and the one that’s more likely to be successful, would be to check with the curatorial staff at a museum or gallery that has a strong collection of early photography, and is likely to have faced these challenges before. Their solution may not ultimately be yours, but at least you may be able to get the benefit of hearing what approaches they considered, what the pros/cons of each were, etc.
Given a damn-the-costs budget and no access to such experts, my approach would probably be to call around until I found a professional photo lab in New York that has dealt with glass negatives successfully before, have them make copy negatives on equivalent-sized film, and then have the negs scanned on one of the desktop drum scanners that are now available; you can get a used Howtek or Colorgetter for between $5K and $10K.
Check around with the folks on Photo.net for more advice.