I use a flatbed UV box that I bought cheap on ebay. It has 6 UV tubes of 9W each. A thick piece of glass (6mm) on the top and a lid with foam to press down on the PCB. One side at a time. After a lot of trial and error and head scratching I finally have figured it out.
1) The UV box must have been on for at least 5 mins prior to exposing anything, and even when turning the board over to do the other side, must keep it on, so the lamps stay warm. Failure to do that means all the timings are off, completely off, one side good, other side bad and so on. This is very important and I bought a lux meter to make sure there is a difference, when warm the lamps are twice the lux output than when cold. Huge difference.
2) To do double sided boards, I print a horizontal and a vertical guide lines on the top left of the board adjacent to the edges, and use those two guide lines to align the board. The reason for only two lines is because the board is not perfectly square. This is very important to note. No matter how it has been cut at home, some sides are better than others (I use a circular tile saw and my cuts are good, but not perfect in terms of absolute squares). I look at the board and choose the two most "straight" sides and mark that corner as "top left". I use this corner to align against the guide lines, if both lines cannot be made perfect, then I choose only one line and make a note which line it is. When I turn the board over, I again align on the same guide line. Using this technique on 110 x 150 mm boards, yields better than 0.5mm accuracy even on the farthest point. I have not yet tried using staples or glue to sandwich the board inside the two transparencies, I cannot see how that will work with exposing one side at a time and having to flip over the whole assembly, and the board inevitably moving.
3) The developer must be at a perfect 23-25C and it takes approx 45 seconds of vigorous splashing to develop. In the winter time the developer must be heated to reach the 24C. a few degrees do make a lot of difference and it is much preferable to spend time getting to the correct temperature than attempting to develop a board which refuses to or over develops. To make sure there is no over development I have 4 very thin crosses on the board - if these disappear it means there has been too much development (or exposure). Typically 45 seconds, max 60 seconds is enough.
4) I have plastic stackable tubs/shoe boxes in which I run all the chemicals, and to heat up a chemical, I put hot or boiling water into one box, and then place the chemical box inside of it (they are stackable). This is the way I do the etching, it takes about one hour, and I re-boil the bottom box water two or three times.
5) My boards are through hole, thinnest traces are usually 0.7mm and gaps are also 0.7mm. I have not attempted better definition.