There should be a version of MathCad for Ipads and the like. Around since about 1985, completely WYSIWYG, and incorporating calculations, graphs, text areas, etc.; and embeded Windows-compatible objects from other applications. Most of all, one can store pages and make topic-specific page templates (e.g., for physics). These calculators came out in the last year of the Apollo program, which I was working on, and they were great then, but obsolete as soon as MathCad became available. Efficiency, discipline and legacy all available in one venue. The academic version is worth every penny if you are going to be any sort of STEM major, and will last you right through graduate school. There is a pretty good freeware clone called SMath Studio available on the web. Any version of StudyWorks from Amazon is nearly as useful, the limitation being that you can't build complex procedures ("programming"), which is almost never required in most lower-level courses.

Well I am (or was) a GPS expert, having participated in the development of the first receiver software. Bcrowell is correct in every detail. As he says the corrections are completely deterministic; the satellite clocks are controlled to run at the rate they would if stationary on the equi-potential geoid, a time-varying correction since the satellite orbits decay away from circular after a while. The special and general effects are comparable in magnitude and first order in the system operation (if not compensated). It was overkill, but I had the honor of coding the platform-dependent relativistic corrections (e.g., a supersonic maneuvering F4): one line of FORTRAN (yes, we had fire in my village in those days). To establish a common time reference at different locations there are various other error sources, measurable (iono delay), or model-based (tropo delay), cables, etc., but I assume these have long since been wrung out. As Newton said "I frame no hypothesis...", but I doubt if the problem is GPS-related. It's too simple.