Focusing on your most profitable products is a solid merchandising strategy.
Sell more of those products and you'll get to keep more money than selling other products.
One hiccup though is that most profitability measurements only look at a time transaction at a time. They assume all transactions are created equal.
You'll quickly realize that transactions involving repeat customers are much better. Not only are their acquisition costs lower (which aren't always included in product profit calculations) but if your customer is satisfied then they are likely to come back and order again. This means there's a chance of another transaction in the future, which has even more profit potential.
This means you should be looking at how a product impacts future customer behavior. The most profitable products from the COGS-view might be different from the most profitable future-customer-behavior products.
The best way to find these products is to measure what customers do in their lifecycle when they buy these products. Measuring how much they spend (LTV or CLTV) is a simple metric that works well. Then you can compare product to product to see which lead to the highest spending customers. Repeat Customer Insights has this product analysis built-in.
(A better metric would be to figure out how much each product in the order contributed to your profit and then measuring that for each customer. That'd be the gold-standard for this but it's a lot more involved calculation and might not turn up anything new over the simple LTV calculation)
When you're figuring out what products to feature in your promotions, pick from both lists. Homepage features, email call-outs, and everything else should promote both types (as well as other types I'm not mentioning here).
You might also lean one way or the other at specific times, e.g. promote the COGS-profitable ones more on new customer landing pages and ads, the future-customer-behavior in your monthly newsletter.
Measure your customer loyalty
Measure the different levels of customer loyalty with Repeat Customer Insights. It uses various models to segment and grade your customers based on their behavior.