I'm not really sure what your question is. You are loading four images into the WS and then you concatenated them into a single image. The images are a single image because you concatenated them into one.
My understanding is that imtile() and montage() can accept a multiframe stack of images (i.e. a rows x columns x channels x frames array), or they can accept a cell array of filenames -- but in either case, the image tiles (the blocks within the output image) must have compatible height and width to be tileable. In both cases, these tools pad or crop the input images to meet that constraint. I doubt this is really what you want. Resizing to meet the geometry constraint is likely preferred.
There are probably various solutions. If you can rely on your image sizes being fixed, you can just tile them manually into a single image. Here, I'm assuming that your geometry specifications are backwards. (i.e. that the images are all 460px tall)
% calculate width of top row
imwidth=size(readMain,2) + size(readTE,2) + size(readLE,2);
% resize the fourth image to match the top row width
% consider the interpolation method used based on the image content
readMid=imresize(readMid,[NaN imwidth],'bicubic');
% concatenate the images
tiledpic=cat(2,readMain,readTE,readLE);
tiledpic=cat(1,tiledpic,readMid);
% show the result
imshow(tiledpic);
If you want padding between the images, you can pad them (e.g. using padarray()) prior to tiling, or you can explicitly generate padding blocks to concatenate between the images. If the image height can't be relied upon to be consistent, you can compensate for any height mismatch in the padding process, that way the padded images can be tiled.
And yes, there's no reason this can't all be done in a loop
