A clamp would draw magnetic flux line into it. This is because a clamp has a high permerability (if it happens to be made of iron, which I would bet it was).
Your plan seems to be ok. You would probabily want to place the two permenent magnets lined up to each other so it produces a dipole and a near uniform field between them. You are right in that you would want to mount the magnets on non-magnetic surfaces. Ideally you would want to find some matterial that has a relative permability as close to 1 as possible. A good example is aluminium, which is a paramagnetic material. I would suggest that the hall probe be attached to a DC current source (a type of power supply that limits the voltage so that a constant current is flowing through the terminals).
By using helmholtz coils you should be able to produce a known flux density by altering the current. Use this to produce a response graph of voltage out vs flux density of the hall probe. Then you replace the helmholtz coils with the magnets, measure the voltage out of the probe, look up the flux density from the graph, and hey presto you know how strong the magnets are.
Hints: The more current that is flowing through the probe the larger the output voltage, however the current will generate heat, which would change the charactoristics of the probe (the Vout is related to the velocity of the charge carriers, increasing the temperature will cause more charge carriers to be liberated if the probe is a semiconductor, thus reducing the velocity of the charge carriers). Keep the probe at a constant temperature.
Since I am bored out of my skull I decided to do the intergration for you on the Biot Bavart Law for a pair of solenoids arranged as Helmholtz coils
B = 2mju0 I / sqrt(10)pi r
Where mju0 is 4pi x 10^-7Hm^-1
pi is erm ...3.141blah blah blah
r is the radius of the Helmholtz coils and the seperation
I is the current flowing through the coils
n is the number of windings of each coil.
This was from the intergration (All bolds are vectors)
dB = mju0 I (dL v a)/4pi |a|^3
a^2 = r^2 + (r/2)^2 (a being the seperation between the wire dL and the point in space we are measuring the flux at)
a = r sqrt(5)/2
Using symetry we can reduce the intergration down to just dL . univector r .cos45
=dLcos45
leaving the intergration being dB = [cos45mju0 I/2sqrt(5)pi r^2] dL
dL is replaced by r sintheta dtheta
Intergrate the whole thing over 2pi and 0
B = mju0 I / sqrt(10)pi r
There are two coils so
B = 2mju0 I / sqrt(10)pi r
Then we multiply by n for each current loop there is...easy...