A CSV-file is not a well-defined and agreed upon format. Most CSV-files have a fixed number of columns. Well-behaved systems should not produce CSV-files with a varying number of columns. What you are describing is a free-format text-file with delimiters. It is not uncommon - but it is quite a hassle to deal with - dirty data.
However you can tell QlikView to disregard any header in the file and force it to read a maximum number of columns. What happens then is that rows with fewer columns will have empty strings or null-values in the missing columns - that will always be at the end. But you might still get unpredictable results unless there is a strict logic to have columns are left out and you can analyze and determine this from some source files.
