I am only suggesting removing one of the radiolocation points because your siphon survey sounds like it's fairly accurate, and as tarquin says not likely to be causing the error. I would think the accuracy of the radiolocation could be the weak link, but i'm no expert with radiolocation.
Depends. I would say something that uses gyroscopes to detect its position, is likely to be quite inaccurate, but that really depends on the quality of the device. I know Andrew has worked with some supremely accurate devices, but in this case, I would consider it quite suspect.
From the specs, it sounds like roughly grade 2 compass (5°), reasonably accurate depth (30 cm), but using gyroscope and accelerometer to measure motion is not exactly accurate, at least in most consumer grade equipment. But I would love to hear otherwise!
If you set the "sd" options correctly for the device, and set the appropriate standard deviations for the radio location fixes, then Therion should be able to sort out the error distribution accordingly. sd for a radio location is likely to be quite large, but actually not as bad as you might expect. Radio location "may be able to locate the horizontal position of an underground point to within just 1.5% of the depth below ground (about 1 metre per 67 metres of depth), and the depth to about 5% of its actual depth, depending on how level the surface is". But that is in addition to whatever error there is in locating the point on the surface where the radio location antenna was used (ie. if you survey to it at grade 5, or locate it with a GPS that has a 5x5x15 m error - fairly good for a GPS, no matter what yours might like to claim). Both of those errors would need to be put into the standard deviation, to get it right. eg. for a passage 200 m below surface: 1.5% * 200 = 3 5% * 200 = 10 3 + 5 = 8 10 + 15 = 25 fix sumpstart 123 123 123 8 8 25 # 8 m easting, 8 m northing, 25 m alt