This is a quick verbal description- sometime in the distant future I'll put up some drawings.
The standard Newtonian secondary mirror holder is a dastardly thing. Three adjustment screws, none of which move the mirror in the desired direction. Interaction between the adjustments, and poor predictability of motion. Yecch!
If you have a small lathe, or are reasonably handy with fabricating things with hand tools, a much better design can easily be built. It's based on the mirror mounts used in optical and laser labs. It doesn't adhere strictly to the rules of kinematic design, but makes up for it in ease of fabrication and real world stability. I'm only giving you the general concept here, so read carefully and make some sketches before proceeding. The idea is to customize this for your situation and available materials.
Fabricate two circular aluminum plates a bit smaller than the obstruction diameter of the secondary. 1/4" thick is about right. Attach a 3/8" threaded rod to the center of one plate. This will mount in the telescope's spider in the usual fashion. Drill and tap a #6-32 hole in the center of the rod on the plate end. (note- if you change materials, keep both plates the same so the thermal expansion coefficient matches)
Fasten the two plates together and drill two holes on the perimeter at 120 degrees. They should be a bit smaller than a 3/16" ball bearing, say 9/64". Place two 3/16" ball bearings between the plates (in the holes) and give the upper plate a sound whack with a hammer. Protect the plate with a piece of wood. This creates a nice seat for the balls on the edge of the holes. All you need is a narrow shiny ring. Drill a third hole at 120 degrees in the plate with the rod. Do not drill the other plate! Tap this hole with a #4-40 tap. Use #6-40 or #6-48 if you have access to that size screw and tap. It will give better resolution. Round and smooth the end of the appropriate size screw and install it in the tapped hole.
You should now be able to adjust the screw and tilt the plates in relation to one another. They should rotate smoothly on the ball bearings. The remaining problem is to attach a spring to hold the plates together. There are a couple ways to do this.
Method 1- Drill a clearance hole in the moveable plate and install a through-bolt and compression spring. This goes in the tapped hole in the rod made earlier. Use a fairly stiff spring.
Method 2- Same as #1 but use a stack of opposed Belleville washers instead of a coil spring. The force will be higher and the unit more stable. Travel will be limited. Probably the best method, but Belleville washers are usually expensive and hard to get.
Method3- Cross-drill a short #6 screw and install it in the rod. Drill a larger hole in the center of the movable plate, ideally with a counter bore. Hook an extension spring on the cross-drilled screw, aim it through the moveable plate, and retain it with a small pin captured in the counterbore.
I should note that a single central spring isn't the only way to do this, but if you use more than one spring, keep them inside the triangle formed by the balls and adjustment screw.
Various mirror mounts can be used. I cut a piece of similar sized PVC pipe at 45 degrees and epoxied it to the movable plate. The secondary mirror was attached to the PVC with silicone rubber (RTV).
Telescope collimation is now easy and predictable. Rotate the secondary holder in the spider as one adjustment. Turn the screw to tilt the mirror as the other. The only remaining adjustment might be to use washers under the post to get the initial mirror position correct.
Enjoy.