Miners look at the sum of sats/byte, and not the sats/byte by itself, right?
Let's say you are a delivery driver. Each morning you go to a warehouse with your truck, and there are a bunch of different jobs you can choose from. Today, you see the following jobs posted:
- Deliver a table and chairs for $50
- Deliver a TV for $30
- Deliver a microwave for $20
- Deliver a toaster oven for $20
You might jump straight for the table and chairs, since that earns you the most money. However, once they are in your truck, there is no room for anything else, and so that's the only job you can do today. By the time you deliver the table and chairs and come back to the warehouse, other people will have done the other jobs.
On the other hand, you can fit the TV, the microwave, and the toaster oven all in your truck together. By taking multiple cheaper jobs, you actually end up with more money than taking the single expensive job.
Think of a bitcoin block like the truck in this analogy, and individual transactions like the various jobs. Although a transaction may pay more overall, it might be better for a miner to pick smaller transactions with a lower overall fee because they can squeeze more of them in to the finite space in a block.