Are your images correctly converted to a binary format? Sometimes ImageJ quantifies a binarized image (225 and 0 values) inversely, based on the settings. The settings may have been changed; it happened to me once. So your negative control would have more signal if the negative areas of the image were quantified and the positive areas were not. Try inverting the image and see if it works. 

I'm not sure about the splitting of the images, but it sounds like it could be a thresholding problem. Try thresholding all of your images following the splitting of the channels to a specific value before quantifying. 

First of all, are the images exiting the microscope JPEGs?

It will be difficult to argue on the scientific quality of your analysis when using a lossy format. If you have these as a multi channel stack, it would be preferable to post them this way. 

Secondly the images are saturated in the blue and green channels, strangely enough the green is saturated at 96-97 so some odd rescaling took place there... 

What Laura suggests can also happen if you are using imageJ to convert your images before measuring. From 16-bit to 8-bit or RGB for example.

It also happens if you are exporting your images as RGB from the microscope that the software is scaling the pixel values. That is suppose your data bins go from 0 to 4095 (12-bit camera), the system will try to fit all this into 0-255 (8-bit range). Some schemes take everything while some others use the minimum and maximum values to rescale the data. In the later case you will have a completely incorrect rescaling on the intensities on a per-image basis, which makes them uncomparable. 

Long story short, could you show us the raw data from the microscope in whatever format it usually outputs it?

I'm not sure how to get the raw data from the microscope, I'll have to ask my PI if he knows how to do that.

However I do save my images as ".tiff ".

This is my current procedure:

1) Split channels

2) Threshold the DAPI channel

3) Use Watershed function from Binary processing

4) Use Analyze Particles function (Set size limit to 200-infinity, do not count cells touching edges)

5) Go to GFP channel and click "show all" on  the Analysis window to make sure the analysis lines up correctly (remove any strange samples that contain multiple nuclei)

6) Create two ovals containing only background to get an average of the background present in the image

7) Measure the particles analyzed and place proceed with analysis to create histogram.

Here are the pictures after they have been split:

1) Negative Control DAPI & 2) Negative Control GFP

3) Positive Control DAPI & 4) Positive Control GFP

http://i57.tinypic.com/k2hf6v.png

http://i60.tinypic.com/kbodom.png

http://i58.tinypic.com/290smu0.png

http://i62.tinypic.com/9t1g0w.png

Looking at your separated png images it looks like image 2 (-ve control GFP) is brighter than image 4 (+ve control GFP).

It appears that either:

1) you have lots of bleed-through of the DAPI signal into your GFP channel, or

2) you are splitting RGB images instead of multi-channel images.

My guess is that you are saving RGB tiffs and splitting them in imageJ. If you cannot get RAW data in the microscopes own file format try saving (or exporting) the channels separately from the microscope software instead.

I will try that, thank you very much!