This could relate to an issue I also ran into, i.e. that some cards from the PCI-PCIe transition period incorporated a bridge to interface between the PCIe bus on the motherboard and the PCI bus on the actual peripheral. These bridge ASICs are inherently problematic in my experience. E.g. some of the "supposed to work" Adaptec "PCIe" SCSI cards (really PCI with an onboard bridge) refuse to work with most PCIe motherboards.
The problem does not exist with native PCI SCSI cards because that's how they were originally made. Cards from the "UW160/320 era", i.e. the later SCSI cards that still used classic SCSI hardware (parallel cables etc.) are often (or perhaps always) PCI cards with a PCIe bridge bolted on, and that's just a nightmare (or: impossible) to get to work.
Have you tested with other SCSI devices for correct functioning?
In my recent SCSI adventures I've made setups that would properly detect the SCSI card, but where the attached scanner really would not work, or not reliably. So correct detection and driver install are not necessarily indicators of a functional setup, although it certainly is a good sign and a prerequisite.
I've had modest success with PCI(e) passthrough on Linux VM's, but ran into trouble due to the issues outlined above w.r.t. PCIe/PCI bridges. But the passthrough as such seemed to work.