I'm not familiar with the kit but I'd to ask you if your protein of interest localizes to the Nuclear Matrix. If so, then it might be really challenging to get good Co-IPs requiring steps like high salt buffers, incubation with DNase and RNase and a bunch of other tricks. If your protein is tagged in some fashion is helpful (FLAG, HA, MYC, 6xHis, etc) or is a fusion protein (say, with EGFP), then there are additional tricks. 

Hi, I think we need more informations to help you.

If I was you, I'd first make sure that your nuclei are well purified. Look for the expression of a nuclear marker (lamin B, histone H3, etc...) and a cytosol marker (GAPDH, tubulin) in your cytosolic and nuclear fractions.

You need to use 2 different antibodies from 2 different species (like rabbit and mouse), one for your IP and one for the western.  If you don't do that, your secondary antibody will bind to your IgG chain like hell.... Also, have you tried a western blot of the whole lysates, to make sure that the antibody is working ? We do immunoprecipitations on nuclear and cytosolic fractions in our jurkat cells, but it is possible that the nuclear extraction buffer from your kit isn't optimal for IP experiments. 

Pierce kits work well.  Oscar and Bruno make strong points.  Start with intact pure nuclei first, then worry about the extract.  Remember, all attempts to extract protein will promote dissociation of your presumed interaction.  Try simple hypotonic/low concentration  non-ionic detergents.  Sometimes Ca+2 helps retain association.  A co-IP in a tube though is a dubious measure of what is happening in vivo.  Do you have a colleague/facility that can do high resolution dual-labeled con-focal fluorescence microscopy and determine a Pearson's Coefficient?

Thank you all for your answers.