I know I'm replying to a question over 3 years old, but perhaps someone else will stumble upon a similar issue and I can help.
This configuration will definitely create a loop. What should happen here is that because it's a LAG, we treat this just like a single connection to the switch. A loop will be detected and STP will shut down 1 of the 2 LAG's, resulting in only using 1/2 the available bandwidth at any given time.
I'd be curious to know more about the STP issue you had. This would normally be fine assuming STP was healthy, but could present less favorable behavior depending on a few things. For one, this creates an LACP link, but LACP is only supported with a VMware Distributed Switch, which isn't included of certain packages.
The best way to go, if possible, is to stack the switches. This usually allows you to span a single LAG across multiple switches since it see's all of the members as a single device. Having an uplink between them usually won't support doing that, so there must be 2 lags, split between switches, and resulting in potentially undesirable performance. But, I'm not an expert on PowerConnect's, so I could be wrong.
It's been quite a while, so I'm assuming a resolution was met. I'd love to know what the solution was.