Have you looked at the load documentation? If you load the .MAT data into a structure, then you can collect all of these together into a (non-scalar) structure.
First lets create a few .MAT files, all using the same variable name A:
>> A = [0];
>> save('temp1','A')
>> A = [Inf,-Inf,NaN];
>> save('temp2','A')
>> A = [1,2,3];
>> save('temp3','A')
Now run this code to load them all at once:
>> D = dir('temp*.mat');
>> S = cellfun(@load,{D.name});
Lets check the third file's value:
>> S(3).A
ans = [1,2,3]
Which is correct. We can also get that file's name with D(3).name.
This code assumes that every .MAT file contains exactly the same fields (but this can be any number of fields). If any .MAT file has different fields, then you can use load's optional arguments to specify what variables to load from the .MAT file:
S = cellfun(@(d)load(d,'A'),{D.name});
