If you want to compare with a particular gene database, your best approach will likely be to map directly to that database. Different gene databases will always have different ideas about what is considered to be a gene (otherwise they wouldn't be different databases). As mentioned a few times elsewhere, translating between different databases is never going to be a 1:1 mapping:

Converting Gene Symbols to Entrez ID in ambiguous cases

• There will be some gene IDs that have no associated Ensembl IDs
• There will be some Ensembl IDs that have no associated gene IDs
• There will be some gene IDs that have multiple associated Ensembl IDs
• There will be some Ensembl IDs that have multiple associated gene IDs

If you can't do that with your own data, or don't want to do that, then you should re-map the GEO data from the original raw FASTQ files.

If you want to compare with a particular gene database, your best approach will likely be to map directly to that database. Different gene databases will always have different ideas about what is considered to be a gene (otherwise they wouldn't be different databases). 

As mentioned a few times elsewhere, translating between different databases is never going to be a 1:1 mapping, so should be avoided if remapping is possible:

Converting Gene Symbols to Entrez ID in ambiguous cases

• There will be some gene IDs that have no associated Ensembl IDs
• There will be some Ensembl IDs that have no associated gene IDs
• There will be some gene IDs that have multiple associated Ensembl IDs
• There will be some Ensembl IDs that have multiple associated gene IDs

If you can't do that with your own data, or don't want to do that, then you should re-map the GEO data from the original raw FASTQ files.