To understand differences in performance in these two cases, you need to consider the hardware involved and the corresponding data path.
To function, a SD (or microSD) card must have a SD controller. SD is a fairly complex standard, and the controller must perform several actions to identify the card capabilities (one-lane vs four-lanes, or even basic SPI) and determine the speed of supported interface. Then switch the data lanes into proper mode (if it supports them).
Now we have:
USB case. You have an internal PCIe bus > [host USB controller] - > USB port > USB "adapter". The USB adapter has a USB-to SD controller bridge, and a SD controller1. Depending on when it was produced, the SD controller may support all latest SD speed modes, or may not.
The built-in SD slot uses most likely something like Realtek RTS5208 controller (like my old laptop has), which resides directly on external PCIe bus. This is a different SD controller, lets call it controller2.
It might seem obvious that the Controller2, being directly connected to the internal PCIe bus should perform better, but it is not for sure. There could be difference in performance in all directions.
First, the SD_Controller1 is different from your SD_Controller2. One may support all speed modes for your particular micro-SD card, the other may stay in older slow modes (since every SD is backward compatible). So the bulk performance will be seriously different.
Second, the USB host is connected to the system via internal 16-lane architecture, which could be running at Gen4 speed. The embedded Realtek likely uses 1x PCIe lane, maybe at old Gen2 or Gen1 speed. This could be a bottleneck, which would favor USB adapter.
Therefore, the performance difference can by anything, all depends on particular microSD capability, SD controller capability, and bus interface throughput.