Mike, I encountered a similar problem in analysing the stained glass dial from Wendon Lofts.
Assuming that there are no signs of the gnomon support to define the sub-style angle on your dial, I would get a value of the declination from the asymmetry of the "earliest" and "latest" hour lines on the dial. Then, having measured all the hour (and half- and quarter-hours if they exist), I would plot y = hourline angle - LHA vs x = LHA. This should give a curve which starts at the origin, rises to a peak, then returns to zero at LHA = 90 degrees (this for the case of a direct south dial - the curve will be offset for the am and pm halves for a decliner). This experimental plot can then be compared to a series of theoretical ones calculated for different latitudes eg by the BSS Sundial Constructor program. The closest fit tells you the latitude of your dial. The advantage of the rather strange-looking x-y parameters that I have suggested is that it (a) it makes use of all the hour line information you have, (b) it is most sensitive to lines around 45 degrees, and (c) plotting the difference between hourline angle and LHA maximises the differences between curves for different latitudes. You may have to iterate around the above loop a couple of times if there is significant uncertainty as to the declination. I hope this helps - I have the Excel plots from the Wendon Lofts dial if they are of any use. John Davis