On getting waterfall data at low frequencies:
Look at your impulse response graph, and notice the part that is marked in red. That is the only data that is used for calculating CSD waterfall (or for non-blended windowed frequency response, for that matter). The idea is that you can window out reflections in the measurement (like in your data, the big reflections at around 2 and 2.8msec) that are due to other stuff in your measurement arrangement, and process only the uncorrupted part that occurs before echoes come in. But, a little math: if you only have T milliseconds of data used, you can't process and get ANY information for frequencies below 1000/T Hz -- you have to have at LEAST a cycle or two of the associated frequency's sinewave (which is longer in time for lower frequencies!) to know anything AT ALL about response at that frequency - otherwise how could it mean anything? Really, several cycles at least is needed to have any confidence at all in that measurement, Omnimic cuts out stuff below several cycles worth at a given frequency (or blends in some echo corrupted stuff, for the frequency response display, if so arranged).
So if you want to get CSD waterfall curves to lower frequency:
1) SET THE WINDOW LENGTH LONG ENOUGH! You set the length by clicking your mouse on the impulse response graph where you want it to end (i.e., go red in color). Tip: look in the rather extensive HELP
in Omnimic, search under "Cumulative", "Windowing", "echoes", "impulse", etc. Or check for similar terms in the HELP
2) If you are looking for a CSD of a speaker only, you will likely have to have the microphone pretty close to the speaker so you get enough time in the impulse response before a reflection (i.e., echo) arrives (you will set the red windowed section of the impulse response to end before the reflection starts). Or be outdoors facing upwards, or somewhere else that can be arranged to get an impulse response without other surfaces for sound to bounce off of.
3) If you are looking for a CSD of a speaker in (and) a room, all reflections are included. Set the window to as far as you can go in the impulse response (you have to use the arrows to make it stretch out further to the right, 250msec or more). And set the CSD graphs to go to much longer time lengths (the part that moves out toward you), you aren't going to have stuff in the room decay in just a few miliseconds, sound takes about 1 millisecond to go just one foot of distance. In room decay stuff (if that's what you want to use a CSD for) it won't mean much unless you have a lot of impulse response time included and let things drop a significant number of dBs (after quite a few sound wave bounces, which take quite a while). You'll also need to include proper smoothing to not see just grassy hash or flat waves.... adjust to taste (i.e., for something that is meaningful).
4) For room decay stuff, a CSD waterfall isn't usually the thing to use. Time Energy Curve (a.k.a. "Energy Time Curve", or ETC) are usually used for that, with selected frequency range and time smoothing values (in the ETC graph tool). Or for something maybe less tricky to adjust, the Bass Decay tool (doesn't require you to do frequency bands separately and tries to focus on just the bass where room reflections are not separable and are necessarily a fact of life).
Did I mention the Omnimic HELP
(It has a search function that works pretty well)