In selected flares that occurred in AR 7260, we have studied the geometry o
f the brightest soft X-ray loop by tracing it on an image. Even under the a
ssumption that the loop is contained in a plane, it is clear that a single
image does not permit us to determine the full geometry. It only provides p
ossible loop shapes as a function of the inclination angle of the loop plan
e with respect to the vertical. However, all the loops that reproduce the o
bserved appearance give the same direction of increasing height as projecte
d on to the image plane. This direction is compared with two relevant obser
vations. Based on 2-D reconnection models that involve a cusp configuration
, it is expected that the soft X-ray loop top source moves upward with time
and that a higher temperature region exists above the loop top. Several fl
ares are found to contradict these predictions, presumably implying the ina
dequacy of the models. Lastly we discuss a possibility of constraining the
inclination angle (and hence the loop shape) with spatially-unresolved soft
X-ray line spectra which are Doppler-shifted due to plasma upflows.