As a heads up, I had written this before all the other replies, and then wound up going out for a long walk and quick grocery run with hubby, so some of the items have already been answered by others ^_^
The reason I suspect you're still getting a good amount of CPU work units from WCG is one of the three checkboxes that comes after the "intermittent projects" section. It's the second one, that says "If there is no work available for the project(s) I have selected above, please send me work from another project.". There is also a dropdown box a bit later under Advanced Options>Processor Usage that says "Allow research to run on my CPU".
stated, World Community Grid isn't known for it's points (most CPU projects in general don't award much in the way of points but they still do a good amount of useful research), but there are a good number of great research projects within it (I run all of them, but my main favourite is "Mapping Cancer Markers", as well as "Fighting Aids At Home - Phase 2")
Regulating GPU load..... Uhm, good question, I usually let mine run full bore so I've never really looked into such, but I do make use of TThrottle to keep temperatures below 75C
With the project count, I say run however many you wish - right now I myself am running 10 projects, 12 if you include one that has no work and one that's non-cpu work intensive, but usually I'm only running five - but do note that your client will only run things as you've set them, and download work accordingly. My main projects are Einstein@Home (my usual GPU project since at present, GPUGrid doesn't take long to ping my GTX 980 up to 80C core temperatures during the summer - can't wait to watercool the card! I run CPU, GPU, and IGP tasks for this project), Poem@Home (another GPU project, but I also run CPU work units even if they don't make as many points), TheSkynet Pogs (aka pogs, a CPU project), Rosetta@Home (for the cancer, HIV/AIDS, and prostate cancer research), World Community Grid (CPU work units, and I let it run whatever sub-projects it will give to me), as well as WUProp@Home (a non-intensive program that monitors what people are running, and how long each work unit takes to complete. It can help people with hardware decide what projects are best suited to the hardware they have).
^Here is an example of what BOINC Manager's Computing Preference window looks like. In this case, I'm running on a 4770k with hyper-threading enabled, so 75% of the CPU's means it will allow up to six threads of CPU tasks (typically six individual tasks, unless you're running MilkyWay multi-thread work units where it will then be one task spread across six threads). I have it configured this way as I also use my processor's integrated graphics card (Intel HD 4600) for Einstein@Home, and through the use of either an app_config.xml file (POEM@Home) or the project's settings (Einstein@Home) I run two work units at the same time on a single graphics card (GTX 980).
You can use the second line "Use at Most X% of CPU time" to reduce the amount of load on your processor. This is typically done to reduce heat and keep a processor (or room) from overheating. I do not know if this affects GPUs or not.
Wow, BOINC Manager looks weird after having used BOINCtasks for a while. These are the screens I'm usually looking at....
I tend to use BOINCtasks because of a few features - one being the ability to easily monitor multiple clients at once (if you're running multiple computers), but I especially love the colour co-ordination. One shade of green for gpu tasks, a different shade for CPU tasks, and the orange indicates the client kicking the work unit into high priority mode due to a deadline being close, or in my case, I downloaded a decent amount of work units and then dropped the maximum amount of work down to three days total, so it's currently freaking out because I have three times the amount of work it thinks I'm supposed to. oh, and the light blue indicates GPU tasks that are currently inactive, so you can easily tell what you have on hand for CPU or GPU work units.
As for the laptop, I say go for it, but like BeerCan said it definitely won't be leaderboard material but science is science. This is a reason why I run BOINC on my two phones as well, even if one of them is just occasionally due to taking it with me when I go to work, go out for walks, etc.
For GPU projects, they will always require a small portion of CPU in order to run. Some projects require more (GPUGrid and POEM@Home come to mind, where they use upwards of an entire full thread), while others don't use that much (Einstein@Home on two threads only uses 45-50% of a single thread). This is why those who are running GPU tasks will make sure to keep a thread or two available specifically for the graphics cards. You can run full on CPU and GPU, but the GPU projects will bog down and take a great deal longer.
EDIT: Oh right, in regards to BOINC Manager and it seeming to do things in a rather silly fashion.... It's because it prioritzes resource share first (default for every project is "100", it can be any value in between 1 and 10k) followed by due date, so when it reaches the project switch timer point (default is 60 minutes) it then decides whether to stay on that current project (as it still retains priority due to resource share) or changes (because of a close deadline on one project, or another project having been determined as now being the one that should be running). Doesn't seem to make much sense up front, but in the long run (usually 48-72 hours at worst) it balances itself out and provides the given share of the available threads to the projects.Edited by WhiteWulfe - 7/10/16 at 9:07am