Another complication is that the closer you get to the poles the fewer high angle satellites you have anyway, so raising the elevation mask by much can really thin them out.
I always recommend that when planning your antenna installation, you should pick a location that has a clear view toward the equator to improve satellite visibility. This becomes more important as you move further north or south.
Ed
Unfortunately in my case my study/lab is on the north side of my house but I moved the antenna to a garden fence with a slightly better view to the south and I seem to get a good number of satellites in view.
I've just recently (3 days ago) raised my mag mount puck antenna by a couple of metres to clear the "gable ended" brickwork of a dormer windowed attic bedroom on the eastward side of the ballasted biscuit tin I'd been using on the flat bay window roof of my 'office' (now my man cave in what is a 1st floor bedroom below the attic bedroom) to stick the puck antenna onto by way of an upgrade on the internally located antenna.
Although this external location afforded a considerable improvement over the window ledge location, reception of SVs to the east was noticeably degraded which was obviously aggravating the positional errors as displayed by the deviation map plots. Since I wasn't too keen on drilling into the brickwork to mount a bracket for even a short lightweight mast whilst standing on a small bay window flat roof some 20 odd foot above our front garden, I decided to try a variation of the original "Ballasted Biscuit Tin Support Pedestal" theme.
Basically, I'd simply replaced the biscuit tin with an old metal drawer ballasted with four small car batteries with an 8 foot aluminium pole clamped with a couple of U bolts into the corner. The mag mount puck antenna then being stuck onto a screw jar lid held in place on the top of the pole using a countersunk screw into a wooden plug I'd hammered into the pole end. Effectively, a "permanent fixture" (courtesy of gravity) that rather neatly avoided drilling into the fabric of the house and any need to step out onto the roof from the attic bedroom window.
It's about a foot short of clearing the ridge tiles but I'd wanted to avoid having the SMA connection between the antenna and extension cables (both 5 metre lengths) dangling just outside of my office window, exposed to our glorious English weather. As it happens, there's only about 8 inches of antenna cable left to spare in achieving my desire to keep the connectors on the dry side of the window opening.
In spite of this shortfall in desired elevation, I'm now seeing a significant improvement in the sky map display which seems to have reduced the magnitude of the deviations being displayed by the deviation map. However, there's still a significant amount of deviant behaviour in the plot of successive fixes around a 'fixed location' which translate into phase deviations of as much as 20 to 30 ns peak to peak[1].
By way of experiment, I tried using an elevation mask of 20 deg in place of the 10 deg mask setting I'd chosen over the 5 degree default. Oddly, this made the deviation map plots worse, rather than better. It would seem that a 10 degree mask with a reasonably clear all round sky view, is my optimum choice. In this case, it looks like SV quantity beats SV quality.
Going from 5 degrees to 10 had dropped the typical SV count from 20 to 18 whilst a 20 degree mask had lost me another 4 or 5 SVs from the plot, leaving me with a count of 13 to 15 locked in SVs out of a total of 22 to 24 detectable SVs. Perhaps an increase in the elevation mask setting can improve accuracy with timing GPS receivers but that doesn't appear to be the case with a common or garden navigation only GPS receiver (NEO-M8N in this case).
I've attached images showing the diagnostic displays in the u-centre application so you can judge for yourself. One thing to note about the SV signal strengths is that they're around 18dB or so down on the original genuine M8N module's signal reports (however, with an active indoor antenna, the signal strength reports are almost identical between the current fake unit and the original, now slightly damaged, unit). I don't think this apparent disparity makes any difference to their performance in regard of the variations of calculated positions in the deviation map display though.
[1] This last source of timing error is essentially a system limitation arising out of the need for the signals to negotiate an ionosphere of varying electron density, aggravated by the varying path lengths due to elevation angle of the individual SVs which can't be fully compensated for by the limited ionospheric correction data carried in the GNSS data packets.
Never mind the the issues of non-integer division to derive a 10MHz or a 100KHz clock from a 48MHz XO or the sawtooth adjustments (complete with 'hanging bridges') used to keep the PPS transitions aligned to the nearest clock edge, it's this last issue of positional fix errors that demand such long (hours' to days' worth of) averaging times to smooth out (at least as far as the use of a common or garden variety of navigation only GPS receiver is concerned).
JBG