Not long ago I evaluated loadui. And my purpose was similar to yours, I wanted to data drive the load test. But maybe LoadUI isn't built for this purpose. After playing around for a while and evaluating whether it was worth the expense of LoadUI, I decided that I could achieve what I wanted using regular SoapUI NG. I discovered that in reality there were two ways to attack load testing.
1. How much load can this thing take before it falls over. (The sexy impressive management statistic and what LoadUI is great for. This can be used to benchmark the system using various requests and monitoring these results release to release may provide insight into system performance. But the impressive stat isn't really useful in its own right. After all, nobody cares if the system can handle 20zillion req/sec if the requirement is only to handle 20/sec)
2. How does it behave in regular day to day load and can I ramp that up slightly to discover how much head room I have to predict system performance in the future using a representative data set. ie, maybe data that actually occurred on a particular day.
I think a combination of these two would be a sensible approach to load testing in my opinion, but I opted for option 2 as a really good way of evaluating the software without going to the expense of LoadUI. I used SoapUI NG and 'replayed' a days worth of data through our system. Now, whenever we have a new release I can run this same test and compare the system performance to last time. I believe that SoapUI is totally capable of doing this is any real world scenario. What LoadUI give you is the ability to vary the load in different ways, ramp it up and down etc, draw graphs, produce reports, but essentially there is no reason why SoapUI can't do the actual service calls