David,
You can solve the problem in various ways. The easiest might be to modify the writer VI by wiring a True constant into the "header (F)" input of your first Write File node. That forces LabVIEW to include a four-byte header that specifies the length of the array you are writing.
When you wire up the reader code the way you did, the length of the array you wire to the "byte stream type " input is irrelevant; by wiring nothing into the "count" input, you are implicitly telling LabVIEW to look for a header to determine the length of the array you are asking it to read. Because you didn't write any header, LabVIEW is interpreting the first four bytes of the array itself as a length. That length doesn't match the actual amount of data present in the datafile, so LabVIEW generates your error.
An alternative solution would be to leave the writer unchanged and modify the reader VI: wire an I32 constant into "byte stream type" and wire a 16 into "count". The choice depends on what you prefer and whether or not these binary files need to be compatible with some other program or not.
Another aside: you can dispense with your case that checks whether the datafile already exists. Just change the function input of the Open/Create/Replace to "create or replace" and wire a False constant into its "advisory dialog" input.
Hope that's clear,
John
