GHartmann makes a valid point re: SG vs power - I almost posted without bearing this in mind. So, I cheated a bit and looked at was scribbled out in the title block to see if I could determine the fluid pumped. It says FLASH DRUM BOTTOMS, which might be low density and low viscosity when hot, and syrup when cold. From the curve, it gives the appearance that 20 HP might be the biggest motor the vendor would plan to put on this pump, so either they or the specifying engineer wanted it that way. For 20 HP, it's not a big cost to worry about. So, except for the uncertainty with SG, you can probably put just about any impeller that you want into this pump.
Assume for the moment - and this is a big assumption - that your suction piping is not undersized so that when you start the pump you don't see a really big drop in pump suction head at flowing conditions. The fact that you mention that you can run two pumps and successfully hit your duty point suggests to me that you are probably OK on the suction side.
To me, what is likely making this pump struggle to achieve higher flow is, as others have suggested, the system head curve. What I would be inclined to do is first read the pressure on the discharge side of the pump with the pump turned off. This should give you a good indication of the delivery pressure (hence discharge head) at zero flow. Next, read the suction pressure with the pump off, which without a gauge on the pump, should be whatever your flash drum pressure is plus the static pressure due to fluid height in the drum. In my mind, this differential pressure (thus differential head) is used to define the point where the system head curve intersects the ordinate; from this point (the vertex) you will have an approximate concave up parabola defined by (K+fL/d)*u^2/(2g). (K is to account for entrance and exit losses; call it 1.5). Irrespective of the ensuing arithmetic, what you might find is that the lower that intersection with the ordinate at zero flow becomes, the steeper will be the parabola near the duty point. What that might signify would be that you might need a significant upsize in the impeller to hit the duty point and, depending on SG, your required motor power will need to be increased accordingly. These effects will not be as significant the higher up the intersection with the ordinate becomes.
In other words, the more of the system head that is comprised of static head and pressure to overcome (as opposed to dynamic head loss), the more difficult it is going to be solve your particular problem without a different pump.
