The OPL chip has a single combined stereo output and this is also how OPL emulators work in genral. As such, there is no way to "extract" the output of a single OPL voice from the stereo mix and send it to a specific plugin. If instrument 1 was assigned to plugin 1 and instrument 2 to plugin 2, there would be no way to know how to "split" the signal because it's not possible.

In the future we might be using a separate OPL chip instance for each instrument in MPTM format which would enable us to work around this limitation, but until then there is nothing that can be done about the current behaviour - it's very much by design.