Comparison of mean signal using RegioProps versus Binary mask

I have an image that I have transferred the results from region props to identify the light fluorescent objects. I want to obtain the mean of this signal. I have attempted two ways and get different answers.

The image showing the result of a regionprops applied to a binary Image

This is the RegioProps way:

%DrawRegions from Binary to raw image:
    labeledImage = bwlabel(Binary, 8);     % Label each blob so we can make measurements of it
    blobMeasurements = regionprops(labeledImage, ROI, 'all');     %ROI is original image
    numberOfBlobs = size(blobMeasurements, 1);
    hold on;
    boundaries = bwboundaries(Binary);  
    numberOfBoundaries = size(boundaries);
    subplot(4,5,[1,2,6,7])
    hold on
    for k = 1 : numberOfBoundaries
  thisBoundary = boundaries{k};
  plot(thisBoundary(:,2), thisBoundary(:,1), 'g', 'LineWidth', 1);
    end
    hold off;

      %Get mean intensity of foreground by blobmeasurements
      allBlobIntensities = [blobMeasurements.MeanIntensity]
      meanI=mean(allBlobIntensities(:))

which gives a mean of 4357 counts.

Using the Binary image used above to act as a mask on the raw Image:

%Now by Binary mask
    IM=ROI;   %Assign original image to IM
    Binary=logical(Binary);
    IM(~Binary)=0
    meanI2=mean(IM(:))

This gives a mean of 2295

??? Have I made an error?

Thanks Jason