Currently the Beaker data model has a single "model" field for disks. IDE disks report just a single 40-char model but SCSI disks actually have separate vendor and model fields (although it seems that very often the vendor value is just rubbish). Separate vendor and model also more closely matches the generic lshw data model for devices, and when we get the disk info from lshw (bug 902567) we have the vendor and model separated already, so we should store it in Beaker that way. There are some backwards compatibility implications which I am not sure about though.
Actually reality seems to be a bit messier than that, the SCSI vendor for SATA disks appears to be often (always?) set to just ATA with the actual vendor in the model field. For example right now we have things like this from parted: ATA WDC WD10EZEX-60Z ATA Maxtor 6L250S0 ATA ST380815AS where WDC = Western Digital and ST = Seagate(?). lshw has a pile of heuristics to discard the useless "ATA" vendor string and determine the actual vendor from the common prefixes. So the right solution is indeed to go with the vendor/model from lshw (and continue to improve lshw's heuristics where necessary to give cleaned-up useful vendor and model information).