I tend to agree with Vodkaman. I'm not a fluid dynamics engineer, but I've seen several of these tanks at icast and other shows. I haven't seen a single concept that I thought made the lure run correctly. I would think the pump will throw out some inconsistent turbulence into the water. So consider this my ecological approach!
In a stream, you will have riffles, pools, and runs. Riffles are shallow with boulders that cause disturbance and create inconsistent flow and turbulence throughout. Pools are much larger and deeper with slower more homogeneous flow. Runs are shallow, but the flow is consistent throughout with minimal turbulence. The substrate in a run is homogeneous and doesn't cause erratic turbulence. So what we need is to re-create a "run"!
In my drawing, flow comes in at a reverse triangle set-up (like what I think Dave is suggesting). This area is open and allows the turbulence coming from the pipe to commingle and (hopefully) reduce. Then water is forced into a constricting point that isn't directly influenced by the turbulence from the pipe. This constriction gradually opens and further reduces the enthalpy of the flow. Then another constriction and the flow comes up into and open area again. Only this time gravity helps to restrict the flow and the flow will "hug" the gradual slope coming up. The red star is where the lure should be tested. To prevent eddying at the tail race, this slope should be gradual too allowing flow to continue to hug this surface to the pipe. It may be more beneficial to have a reducer going from the tank to the pipe on both ends too. Anyway, this idea will more than likely fail... but it's the approach I would take. I think the larger the set-up the better the results.