Mapped onto pseudogenes?

[yumiya]

I have a problem in TopHat-Cufflinks pipeline. I want to get splicing information of a gene family, but the annotation is bad so far, that is, nothing is annotated or huge exons are annotated.
This family has a number of processed pseudogenes with almost same sequence. It seems that reads that are supposed to be mapped onto "real" genes are mapped onto these pseudogenes.
Now, I have a gff file defining the location of those pseudogenes in the genome. Is there a way to mask these pseudogenes in TopHat mapping?
I do not think -M option in Cufflinks works for me, because the reads are already mapped in .bam file from TopHat.
I really need help. If someone knows, please tell me

[yumiya]

Masking

I masked pseudogenes on the genome with "NNNNN....".
That worked at least partly.

[Dario1984]

Some pseduogenes are transcribed, though. So, this isn't a good solution, either. It's a problem for us, too. We get lots of poor transcript assemblies because of this.

[yumiya]

Yes, that's true, but in our case we know that expression of pseudogenes is at a very low level compared to real genes. Moreover, our goal is qualitative rather than quantitative, that is, identification of alternative isoforms.
If I find alternative isoforms in some genes, I can go back to the masked pseudogenes to see whether they might give the identified isoforms (I am talking about "processed" pseudogenes).
However, masking is of course not a good solution for quantitative analyses...

By the way, I found another key. If you are using TopHat and your target genes are highly expressed, --initial-read-mismatches should be 0. Minimize loss of reads derived from true genes by pseudogenes in initial mapping with this option.