Hi Nik,
Usually with hydrous phases, you need to recalculate the expected H2O content based on mineral formula (i.e., assuming a fixed atom per formula unit of OH-group). Such a mineral formula recalculation can be also helpful to check the site occupancy and possibly identify "bad" analysis (i.e. how much cation should be is site A, B, T... and compare with the ideal case). I'm sure you can find on the web or ask a colleague. I also have a website you can calculate mineral formula, and there is a specific entry for amphibole (
http://cub.geoloweb.ch/index.php?page=mineral_formula&reset=true). You can also output your results in Probe for EPMA as atomic proportion and recalculate these values based on a fixed amount of cation (e.g. 15 cations for amphibole, with 22 oxygens and 2 OH group), but I don't think there is an option there to recalculate the H2O wt-%... I also attach a spreadsheet for amphibole mineral formula calculation. You should also check the literature for amphibole classification (e.g., papers from Hawthorne 2007 [Reviews in Mineralogy and Geochemistry 67, pp. 55-88] or Leake 1997 [American Mineralogist 86, pp. 1019-1037]), I think these papers contain useful info on the variation of amphibole composition, which could be of some help for your issue, along with some help about mineral formula calculation / classification of the amphibole.
Once you have your mineral formula recalculated and you have an estimate of H2O content, you can then check for outliers based on the classical assumption that a "good" analysis should be between 98 and 102%. If data are below 98%, something is wrong. However, the mineral formula recalculation can be tricky for these nasty "garbage" amphibole for numerous reasons. Here are some...
a) F and Cl can substitute for OH (did you measure them?). The amount of F and Cl will have a direct influence on how much H2O could be present. There can also be some O2- present instead of OH-...
b) Fe is both Fe2+ and Fe3+ in amphibole, and you cannot reliably measure the Fe2+/3+ proportion by EMP. The proportion of Fe2+ and Fe3+ will influence slightly your total as there is more oxygen associated to Fe3+...
c) vacancies exist in amphibole. The presence of vacancy is problematic essentially when checking the site occupancy.
d) Are you sure you have analyzed every single element in these amphiboles? Any missing element will induce a slightly lower total, and might screw up your mineral formula recalculation. However, if you did analyze all elements you can see on the EDS, then your total (with H2O recalculated) should be close to 98-100%. For minor / trace element you could do a long WDS scan. However, in most cases, the simple analysis of Si, Al, Fe, Ti, Mn, Mg, Ca, Na, K, F, Cl and possibly Cr, too, should be enough...
e) Totals might be too low due to, for instance, a bad standardization of some elements or the alteration of the amphibole (e.g. chloritization => more OH-group than normally expected...). A way to know if the problem could be sample or instrument / calibration related is to analyze a secondary standard. Any chance you have a Kakanui hornblende or another hornblende standard (reference material) of known composition? If so, try to analyze such a standard, and compare your analysis with the certified values...
Good luck!
Julien