I had the same problem and nothing could solve it. Plus, the synthetic backup process was taking much longer to backup my servers due to the initial scan for changed files/folders. True image has the same problem, so I couldn't use it either. Really disappointing. Then there's the very useful "duplicate to tape" feature, except that you cannot use it with true images, which are required when backing up Exchange stores with the "brick level restore" option enabled.
Synthetic incremental backups refuse to work with device pools, even if all of the devices contained within are disk-to-disk folders. Why couldn't it just figure out what each of the devices are and make sure to use only the disk-to-disk folders within the device pool?! All you then have to do is fail the job if no disk-to-disk folders are available for use in the device pool.
Disk-to-disk works fine, unless BE needs to create IMG disk folders, such as the case for Exchange stores with the "brick level restore" option enabled. In that case, it cannot properly figure out how much disk space is available for use by the disk-to-disk folder and ends up running out of disk space and failing the job instead of simply jumping to another available disk-to-disk folder within the device pool. Where's the intelligence in that?
BE has so many cool and great features, but you cannot use any of them due to catch-22 issues or usability issues. :(