"Why aren't d and dT transposes of each other?"
Comma-separated lists are a powerful and useful feature, once their concept is understood.
The reason why is because the comma-separated list that you generated is not one variable as you seem to think (which can have a size or orientation) but is simply a list of scalar variables, and such a list does not have anything like a size or orientation. It is just a list, separated by commas. Both of your examples produce exactly the same comma-separated list:
1,2,3
and then call the cell array constructor with that (same) comma-separated list:
{1,2,3}
The size of your output (if any) depends entirely on the function that you are entering that comma-separated list into as input arguments (i.e. the cell array constructor). Of course with matrices/arrays the order of the comma-separated list can be different, depending on the orientation of the array, but for vectors their orientation makes absolutely no difference.
So, simply put, a comma-separated list does not have size, therefore the output size depends only on the function that you are using that comma-separated list with.
Read these to know more:
