Dave
In general the high throughput math should be used when the normal throughput math is not fast enough. Each function in LabVIEW FPGA will be associated with a certain time delay which will vary depending on the compile. Also some math functions are not supported in a single cycle timed loop (like divide). So when normal math is not fast enough we use High Throughput Math, a library of highly optimized functions can maximize the overall throughput,(even if it adds delay) for each node. When working with high throughput math you have to be more aware of latency and delay but in the end it will allow you to get a much faster output.Check out the following topic from the LabVIEW help for more info on the caveats.
Placing the High Throughput Math Functions in a Single-Cycle Timed Loop (FPGA Module)
I think the math error you are seeing is likely caused by latency timing issues or data type. Take a look at the linked help doc for more info.
