I apologize, but I'm finding it hard to trace what exactly this script is doing. I did take a look at the first failure and sure enough the record leader says it's 463 bytes but the record itself is 464 bytes. So a failure is warranted -- given the current behavior of MARC::Record.
Perhaps dumbing this test script down a bit and making it clear what the heck is being tested would help (at least this developer). //Ed