[Genome] short results when running my own blat server

[Alex Holman]

Hi,
I'm trying to run a local copy of the BLAT server so I can hit it  
with a ton of search requests to the human genome.  I downloaded the  
compiled blatSuite for OSX, and have downloaded chr20.fa from the  
latest alignment of the human genome for testing.  However when I  
take one line randomly from chr20 (50 bases), paste it into a fasta  
file and use it to blat search against chr20.fa, I only turn up a 37  
or 38 base hit.  If I use the same sequence in the web blat at UCSC I  
get the full 50 base hit.
My command is this:
./blat -t=dna -q=dna -out=pslx ../genome/chr20.fa testseq.fa output.psl
though it doesn't seem to make any difference if I specify the target  
and query types or if I use the -fine option or set -minMatch and - 
minScore to 0.

Searching with 100 bases of sequence I get a full length length hit.   
What settings can I change to better replicate the sensitivity that I  
see with the web based tool?

Thanks
-Alex Holman-