I wouldn't let Datacolor off the hook all together here Mark. They did design the instrument and process so if a number of users are experiencing what you and the OP report, that's bad. It should at the very least report a questionable measurement or, like the EyeOne and Munki, report errors to the user. Bad enough it requires a one patch at a time process which sounds like agony (from someone that did this back in early 1990's using a ColorTron <g>) but to end up with a lousy profile is kind of inexcusable.
Well, there is truth in what you say. The Spyder3PrintSR does have a scan mode that, once you get the hang of it, is reasonably convenient. But, its "scan" mode is certainly more primitive than the Xrite method for the i1 and Colormunki. As I understand it, the Xrite instruments take multiple readings as the unit moves over each patch and then Xrite uses the software and reference values to isolate the "good data" that is read somewhere within the area of each patch. Spyder3 printSR uses a very basic IR detection scheme that looks for the black/white stripes at the leading edge of each patch which tells it that the next patch is coming. Then it further relies on a simple timing mechanism (something like 0.5 seconds after the IR stripe detection) to trigger the color reading of the patch.
The obvious weakness in the Spyder3PrintSR scan mode method is that if you are moving the instrument at a steady pace forward but nevertheless going too slow or too fast, the single color reading doesn't occur in the middle of each patch. The software does detect when you clearly miss, but it doesn't easily detect a situation where, for example, the aperture has not made it fully onto the patch and a little of the stripe is still being integrated into the reading. Hence an error, but not so big that the software can detect. Ditto, if you move to fast and the reading takes place with the aperture now starting to hit the trailing edge stripe. Thus, for users who haven't taken the time to figure out this end-user scanning pace like I did, the Datacolor product does invite some user-skill problems. With the bigger "EZ targets" there is more forgiveness, with the smaller "Expert" targets, I consider this scan mode technique to be challenging at best. Better to read the smaller-sized Expert patches with the single-patch reading mode rather than the new "scan" mode.
Also, if you follow my logic on how the Spyder3PrintSR detects patches, you will conclude as I did that a way to improve one's odds of landing in the middle of each patch at the right time is not to go at a steady scan speed forward, but rather to move quickly over each stripe then pause deliberately in the middle of each patch and wait for the "click" sound taken as it measures, then rinse and repeat as they say. This quick-slow scan technique improved my measuring success greatly, and I was even successful with this measuring technique on the smaller Expert patches. Anyway, I suggested to the Datacolor team that they build into their software a feature like Xrite's Measure tool that let's you compare two sets of readings. Then offer the ability right in the software to go back and remeasure the patches where the two readings don't agree). I was told that this feature would be too complicated for the typical user, but ironically, the current Spyderprint software does support advanced features like averaging of two data sets and exporting data to text files. Go figure!
So, to be fair, the Datacolor product does build decent profiles, and has many nice features in the software that photographers will like, but it also demands some skill and patience from the user that some people may find frustrating. Simply put, make sure you are are getting the instrument fully in contact with the calibration tile by putting a little additional downward pressure on the head, and also take the time to train yourself on the correct pace when using the instrument's new "scan mode" feature. Once I had these two issues figured out, then I was able to get very good and repeatable results. That said, I heartily recommend the "compare two sets of measurements" routine for optimum results, but this perfectionist's approach does mean having to take two complete sets of readings, and also having some knowledge of Excel since you can't do the delta E comparison in the software at this time.