Unless the table is just huge or there really is never to be any need for the other variables, there's no reason to not keep the original as is -- or, you could set the import options object to only read the wanted variables first. But, that's an aside; to select the company of interest since it is data inside the variable you have to to a comparison test; it isn't a variable that can be named --
ixPayPro=matches(T.Odbiorca,"PAYPRO S.A."); % the indexing vector
TPayPro=T(ixPayPro,:); % make a specific table of those only
BUT...this really isn't the way to use the table -- instead, treat the variables of interest as grouping variables and operate on the table with them as described in the overview <grouped-calculations-in-tables>. You've not given a specific analysis of interest here, but rarely is actually breaking the table up the right way to proceed.
