To answer "anyone know why" first, take a look at the README in FTP genomes which includes this line:

Plasmids: sequence and annotation of RefSeq plasmids

So only plasmids with RefSeq accessions are found there; you have GenBank accessions.

I'm pretty sure that you will not find a separate GenBank division for plasmid sequences in the FTP site. If you don't want to mirror GenBank locally or use eutils, you could try Batch Entrez, then Send to -> File -> Fasta.

Your best bet is probably eutils. You can retrieve 500 seqs with every call so even several tens of thousands isn't such a big deal. Here $1 is a list of GIs and Entrez Direct is assumed to be in $PATH:

#!/bin/bash
split -l 500 -a 5 $1 input.
for f in input.*
do
IDs=$(cat $f | tr "\n" ",")
epost -db nuccore -id $IDs | efetch -format fasta > $f.output
done
cat *.output > plasmids.fna
#Everything worked out? rm input.* output.*