I can't say I entirely understand how the Elapsed Time Wait works, but testing in qemu, I was able to get it to a good speed by replacing the

  inc cl

with a

  nop
  inc dx

By opening in a hex editor, finding

  bf86 fec1

and replacing with

  bf86 9042

If you're on a very slow system, maybe just use 2 nops or a 2-byte nop.