Notice that as the joint clearance gets tighter and tighter (moving from right to left along the bottom axis), the strength of the joint gets higher and higher (moving upward along the left-hand axis). There is a lot of experience with this over the years, and general acceptance of this information is widespread. However, something strange happens at the far left of the chart in the area represented by gap-clearances of about 0.0015 inch (0.04 mm) or less. There appears to be a drop-off in strength of the brazed joint.
I heard a PhD metallurgist tell an audience during a conference some years ago that people should not allow their brazements to have gaps tighter than 0.0015" (0.04 mm) because joint strengths weaken when they are tighter than this. I was amazed to hear this erroneous information, and quickly realized that probably many people, in the absence of complete information, might assume the same thing.
The original Handy & Harman report is apparently no longer available, but the data for this chart was generated by flame-brazing (torch-brazing) two pieces of 304 stainless sheet stock together in a butt-joint configuration, using silver-based brazing filler metal (BFM) along with a brazing paste-flux (since it was being brazed in air with a torch).
The test pieces being brazed were apparently designed so that the cross-sectional width (and thickness?) of the stainless on each side of the joint was much greater than that in the area of the braze. Thus, the test specimen was tapering down as it approached the joint area so that failure would always occur in the joint and not in the base metal. Thus, the increased values of "strength," shown in the chart, represent the "strength-to-failure" (tensile strength) of the BFM itself and not that of the "overall joint," which would include the stainless, etc.
It is interesting to note that the tensile strength of the silver-based BFM itself is raised by the constraints of the proximity of the sides of the joint. Thus, a silver-based BFM that might have a tensile-strength of about 40,000 psi were a rod of that material pulled apart in a tensile-testing machine, when that same BFM is melted into the confines of a brazed joint the tensile strength of that BFM - in the joint - is modified by the constraints of faying surfaces on each side of the gap.
As the gap-clearance gets tighter and tighter, the normal mode of metallic deformation along preferred slip-planes can no longer effectively take place since the gap is so narrow and actual molecular-bonding rupure-mechanisms enter into the picture instead, requiring far higher levels of force to break the joint. Thus, the chart shows higher and higher "strength" levels for the BFM, up to more than three times the levels of force required to break the BFM in non-constrained rod-form out in open air.
Flux causing drop in the strength-curve below 0.0015"
Remember that these test pieces were brazed in air with flux. All brazements joined in air using flux WILL (not maybe) contain some flux residues. There is no such thing as a flux-free joint when brazing in air with a flux. Thus, when the test pieces were being torch-brazed with gaps at about 0.0015" (0.04 mm) or less, the inevitable flux voids began to become a noticeable part of the brazed joint, and began to negatively affect the joint strength due to their increased presence (percentage wise). Had the joints been able to be "wiped" (surfaces moved back and forth relative to each other while being heated with the flame) to help remove some of those voids it might have helped, but that was apparently not done.
The drop-off in strength values on that widely used chart thus has nothing to do with so-called "negative effects of close joint clearances" but only to do with the inevitable presence of flux (and thus flux voids) in the joint. Were these parts to be brazed in a protective inert atmosphere such as nitrogen or argon, there would have been no fall-off of strength values. Instead, they may have continued to rise a bit more on the chart.