I must say I'm a little skeptical of tests run by a HD manufacturer, especially when the results make a competitor's products look bad or would make you want to buy the manufacturer's significantly more expensive hardware line.
Things to consider: if you extended the test out further, you might see that the failure rates even out (eg, pushing the drives hard killed the 'bad eggs' sooner, but they would have failed in another few thousand hours anyway even at a 'desktop' workload.) They also didn't (at least in this summary) show comparable data for SCSI disks -- maybe they also fail 2-4x as much when pushed with a hard workload than a 'desktop' one. And no definition is given here as to what a 'desktop' workload consists of. If they're doing something like having the drives sit idle 50% of the time, any drive would fail far less often running that workload than running a 100% random read/write one. And what were the environmental factors -- were they pushing the drives at high temperature/vibration to induce early failures?
Generally speaking, most 7200RPM SATA/IDE drives are going to be about the same in reliability (barring a few notoriously bad product lines, like the IBM "Deathstar" Deskstar drives from a couple production runs ~5 years ago.) Some will work fine for years, others will fail. Buying expensive SCSI or FC drives will probably give you a little better MTBF, but you'll still get bad drives from time to time. MTBF numbers (and even studies like this) only give statistical data -- the actual amount of time a given drive will run before failing has a huge variance, and there's no way to definitively tell in advance.
If your issue is data security, make regular backups no matter what kind of drives you have. Buying SCSI doesn't make you immune to needing backups.
If you care about uptime, it is far better to buy two cheaper SATA drives and run them in RAID1 (or 3+ and run RAID5/RAID6) than to buy one SCSI drive, even if the SCSI drive is actually more reliable than either of the SATA drives by themselves.