Can you see by your eyes the difference between biofilm and resin?

Also it would be helpful if you can post image where the problem could be seen.

Have you have done a control staining? i.e. Stain the resin only to determine if it truly picks up any stain. Also to have a sample of bacteria that is unstained to determine the base levels of any background fluorescence. I had some experiences where my background was very high. Alternative is to consider another type of stain? Or a biochemical assay? Hope it helps..

I would advise to introduce a fluorescent protein instead of this staining. The problem is that free DNA is part of biofilm and PI will bind to it, increasing the background but also Syto 9, not only to the resin. What you can do is to treat the resin alone with PI and S9, observe the sample in the CLSM and adjust the exposure (of the camera) and gain to the minimum of fluorescence, this mean to have almost a black screen (equivalent to erase the background). Then use this settings to observe your samples. Any additional fluorescence should come only from the sample. Hope this help.    

@Alexandro: Your suggestion of using resin alone treated with PI & S9 as a baseline for detection of fluorescence from the biofilm sample would not provide an accurate measure in this instance, unless one is able to accurately quantify the binding efficacies and concentrations of PI & S9 bound to the sample and the resin.

@Shelley: Have you tried reducing the pinhole size? Proper occlusion of out-of-focus emission frequencies below the focal plane is what is required in this instance, so that only fluorescence detected in the focal plane of interest (corresponding to the biofilm) would be detected; background fluorescence emanating from the stained resin would be excluded.

You might try TTC staining where metabolic active bacteria will convert the colorless TTC into a insoluble red compound. This method is kind of specific. You can measure the fluorescence. Or You can extract the red compound by EtOH/Acetone and measure by absorbance.   

http://www.ncbi.nlm.nih.gov/pubmed/25086178

http://www.ncbi.nlm.nih.gov/pubmed/23910098

Have you tried fluorescence in situ hybridisation (FISH).  You just need a specific complementary RNA sequence for S. mutans that has fluorensence probe eg. 555.  This probe will definitely bind specifically to S. mutans.  I've done this and it worked with my flow cell biofilm of C. albicans, S. mutans and A. naeslundii, both mono- and polymicrobial biofilms.  Unfortunately, you can't differentiate between the live and dead cells but you can still use COMSTAT to enumerate the cell thickness, biomass and etc.

Hafiz proposed a useful initiative of using FISH. Just to add to Hafiz's suggestion - you can use FISH in conjunction with the S9/PI stain that you are currently using, then employ quantitative colocalization microscopy (qCM) to identify regions where both fluorescence from FISH and S9/PI are detected - these ROIs represent your cells. Regions where only S9/PI fluorescence is detected represent the resin (or background fluorescence) since the FISH probe is not bound to the resin (you'd need to compare the correlation coefficients gleaned from qCM in this respect though).