trouble indexing fasta database

[Gene Cutler]

I am trying to index a fasta database and then retrieve sequences with seqret.
I used dbifasta on half of the human genome golden path from ensembl:

dbifasta -idformat simple -dbname ensembl_gp1 -directory . -filenames 
'ensembl_golden_path.1' -indexdirectory /scratch1/databases/emboss/

and put this in my emboss.default:

DB ensembl_gp1 [
         type: N
         method: emblcd
         format: fasta
         dir: /scratch1/databases/raw/
         file: ensembl_golden_path.1
         indexdir: /scratch1/databases/emboss/
]


If I try to fetch a sequence located in the beginning of the fasta 
file,  it works fine.  But sequences further into the file aren't 
retrieved:

>  seqret ensembl_gp1:AC007798.1.15164.16328
Reads and writes (returns) sequences
    An error has been found: EMBLCD Entry failed
    An error has been found: Database 'ensembl_gp1' : access method 
'emblcd' failed
    An error has been found: option -sequence: Unable to read sequence 
'ensembl_gp1:AC007798.1.15164.16328'
    There is a serious problem: seqret terminated: Bad value for 
option and no prompt

However, when I tried retrieving all the sequences:
>  seqret ensembl_gp1:*
I got them all back.

I can also retrieve the sequences if I get them directly from the 
fasta file rather than using the index.  That makes it seem to be an 
index problem.  Recreating the indexes hasn't helped.