I'm not sure how my version of Anaconda compares to yours - I'm working with a recent test build with a git master updates.img on the top - but I think some of your concerns are addressed in the UI, at least the version in git master:
- It's true that depending on the options you pick (e.g. mirroring) that you can't always specify the exact size of a given partition: we ask users to specify their desired disk capacity, so if you put you want 100 GB on a partition and select mirroring, that partition with be 200 GB in actuality. But there is a way to specify size - does the fact that the field is for desired capacity cause you any issues?
- Per partition, you can specify which physical disk a partition lives on. Select the partition you want to move, then click on the gear icon (older builds) or the screwdriver icon (newer builds) on the bar at the bottom right of the partition list. It should pop up a dialog that asks you which physical disk you'd like the partition to live on. (It should look something like this: http://linuxgrrl.com/fedora-ux/Projects/Anaconda/Prototyp... )
- I know initial versions of the new installer didn't support working with pre-existing partitions so well, but I'm pretty sure that's either been fixed or in the process of being fixed. The engineering team is definitely aware of it and I've seen discussions between them and QA about correcting that issue, so it's definitely on the radar and the fact it wasn't working was not intentional (I think it was more that it simply wasn't finished.)
- Re: "standard nomenclature like /dev/sda and /dev/sdb to identify drives" - the problem with relying on strings like this to identify drives is that they are not guaranteed to be consistent, especially if you physically move drives around but also depending on the order in which the system recognizes the disk. We try to instead provide more reliable identifiers such as the FS label, mount point, the disk manufacturer data and serial number, etc. - although when we can't get any other useful information on a partition, we will be displaying the /dev/sd* nomenclature as well. I know that we're going with a bit of a new way of representing the partitions here - the idea is that users would want to see how their installed system would look primarily, and the hardware and partition-level data and options is more secondary. It may prove to be the wrong approach but we did a lot of mockups and design changes (I think there have been over 100 iterations just on the custom partitioning screen design) and talked to long-time users and gathered feedback on them, and we didn't find any huge red flags pointing to this approach being fundamentally flawed. If it does end up causing a lot of problems for users though, we will definitely re-evaluate it.
I mentioned in another comment that we're planning to do a usability test of the installer after the Fedora 18 release - we'll also be reaching out for even more user feedback at that point, so if you're really stuck please let me know. We really want to improve the installation experience for all users - the redesign process has spanned almost 2 years now starting with the storage filter UI design, we've put a lot of effort into this but if it's not working for users then we've goofed, and we'll try our best to get it right.