SD card's speed isn't constant. It depends on the workload and the condition of the NAND flash memories inside the card. Manufacturer would only put sequential read/write speeds, usually on bare-metal testing platforms, on the package. As there's less free space on the card, and as the card begins to wear out, the speed would decline, sometime rapidly, especially when the card needs to perform background GC or wear-leveling for its SD cards. Testing it with a full OS system would also decrease it's throughput.
By the way, bigger size usually means the NAND chips were made by a more advanced fabrication technology node. While "advanced fabrication" may sound good, it what it really means is that each memory cell is made smaller, with less capacity to store electrons, making it harder to write and more prone to data loss. "Larger capacity" usually means instead of storing one bit of data on one memory cell, now they are storing 2 bit of data, with a technology called MLC. MLC also makes it harder to write and more pron to data loss. So to make sure users don't notice all that data corruption, they added ECC, to repair all the corrupted data so you wouldn't notice the "advance" of technology.
Don't confuse me with your reasonableness.