Yes, re-read the text file and ignore the results of the PROC IMPORT. You can read them one by one or try to use the trick on the other example data step to read all of the files at once into one large dataset. To read all of the files with one data step they must have the fields in the same order on the lines. So if the structure has changed over the years you will want to read them one by one (or group the like files together anyway.)
If you write your own data step to read the text files then you can be assured that the variables are all created exactly the same. This means that you can combine the dataset if you create them separately. It also means you can prevent variables from being define with the wrong type (numeric or character) or wrong length.
You can use the results of the PROC IMPORT to help you figure out how to define the variables in your program. But you probably will want to refer to the definition of the variables as published on the source website. Again some of these fields appear to have only digits in them but they are not NUMBERS. They are identifiers. It will probably be useful to define such variables as character. That is the problem with using PROC IMPORT. A text file has no place to store the attributes of the variables so PROC IMPORT just guesses what to do based the values it checks.