If you have trouble using different people, there is no rule that you MUST use the person category in your system. You could just build an AO system (Action-Object) and then when you memorize, imagine YOURSELF performing the indicated action to or with the object.
This has the potential to provide really strong links, especially if combined with a memory palace sequence, because instead of you trying to recall “ok, at the couch, who was doing what with what?” it becomes “ok when I was at the couch, what did I do?” It provides an opportunity to engage more senses in the encoding process, instead of just observing “michael jordan eating a safe” you would recall yourself trying to eat a safe, the weight of it as you held it, the metallic taste and the crunch of your teeth breaking as you tried to chew it…
Your SCENES would only encode 4 digits, but the important intentional elements that you’re using to encode data would still be encoding 2 digits per element. The same efficiency as a PAO where it really matters. (More on that here and here.) You would need a few more loci to accommodate the full sequences if they’re longer, but that is almost a non-issue for most and a trade off that could be absolutely worth it for scenes that are clearer, simpler, and easier to recall.
If you try that approach, I’d advise basing your elements on Major so that you can more easily develop independent associations for each element. Or at the very least, base your objects on them.