I'm not seeing how any of the statistic columns would capture a value that doesn't actually appear in the table...(actual ... row=0)
Unless there is some prefix matching going on here since the short value is a substring(1, n) of the longer one which does appear 5 times.
I guess maybe because the value doesn't appear it uses the index (via IOS) to confirm absence (or near absence, i.e., 1) while, knowing the larger value appears 5 times out of 223, it decides a quick table scan is faster than any form of double-lookup (whether on the visibility map or the heap).