VoxForge
The pronunciation dictionnary is HTK specific and is different than the file you created in Step 1. The HTK file is used in creating Acoustic Models. The sample.dict file is used in Step 1 is part of the Julian Grammar. The Julian sample.dict file will usually only contain a subset of the words and pronunciation information contained in the HTK pronunciation dictionnary.
Note: There is duplication of pronunciation information in the Julius sample.dict file (part of the Julius Grammar) and the HTK pronunciation dictionary (used in the creation of HTK Acoustic Models). This can cause errors if you don't get your pronunciation data just right - so be careful.