The idea that all early reflections below 15ms are extremely bad for the stereo image and inevitably lead to sound coloration is often found in forums.
Sometimes it is also demanded to suppress the reflections up to 15ms completely.
it's really dependent upon end-state response and room requirements. a studio (control/mix/mastering room) has requirements to be an objective, critically-accurate reproduction space where mix/mastering decisions can be made without the bounded acoustical space masking or imposing distortion or changes in perception with respect to the direct signal - and thus where said decisions translate to a myriad of other reproduction environments and sources. ie, it is to be an objectively "neutral" room.
the Inter Signal Delay gap (ISD, redefined from Manfred Schroeders "ITD" from concert hall research) - is an effectively anechoic time-period where-by only the direct signal is allowed to be processed, and is usually defined from 15-25ms in length. other constraints (such as a tracking room ISD, which should be 2-5ms shorter than the control room's ISD) may be the primary constraint on this value, but not always. the absesnse of a tracking/recording ISD does not imply the reproduction room's ISD is no longer relevant.
what's missed in many of these discussions is that this defined ISD defines the psycho-acoustic size of the bounded space. a smaller room that imposes earlier high-gain first-order reflections will be perceived to be a "smaller space" than that of a room who imposes first indirect energy at a later point in time. therefore, delaying in time (via absorption or redirection) when the first indirect energy impedes the listening position will impart perception of an acoustically larger space than the room's natural boundaries allow. this, from a reproduction standpoint, will also allow one to "hear more into" the tracking/recording room (or digital reverb fx) of the source material than would otherwise.
we know from toole's research that loudspeakers with poor horizontal off-axis response should have sidewall/lateral sparse indirect first-order specular reflections attenuated (ie, with broadband treatments such as porous absorber) - and for those with good off-axis/CD polar response, he states "it is a matter of taste" whether one prefers a (to some) subjectively pleasing apparent source widening (at the expense of image accuracy) or to absorb if one has requirements for an objectively accurate reproduction space in terms of imaging, localization, and intelligibility.
but the key is the loudspeaker characteristics in terms of polar response and the impedance of the boundaries - something that is rarely defined at the start of the conversation and of which can help decide someone of "what to do".
I think that a specification that is intended for studio recordings was transferred to the hifi world. For studio recordings the requirement is to avoid early reflections up to 15ms, because otherwise the recordings are "blurred" (without an increase in spatiality).
This is somehow plausible, since early reflections always occur when listening in living rooms, it is certainly not an advantage if the music signal contains additional early reflections.
typical home residential living rooms are naturally (physical and acoustically) small. the first-order reflections arrive 1) earlier in time and subsequently, 2) higher in gain. attenuating these first-order reflections is a relatively benign and effortless way to increase the perceived size of the reproduction space - and thus allow the ear-brain to more easily hear into the indirect sound-field present within the recording (if applicable).
this will also have positive effects on the frequecy domain (low-mid band), as the higher magnitude reflections constitute to larger peaks/nulls within the subsequent comb-filter interference pattern generated by the superposition of the direct and first-order reflections as viewed within the frequency response.
A very important aspect of early reflections in living rooms is their masking. The negative effects of early reflections like comb filter effects or roughness and beating are masked by other reflections.
if you want to "fill in comb filter effects", sparse specular reflections from bare sidewalls are not optimal. instead, diffusers (such as reflection phase gratings) will convert single sparse reflection into many reflections both spatially and temporally dispersed. the magnitude variations of the comb-filtering thus can be greatly reduced, because of the phase difference of the array of delayed reflections (due to varying depths of the wells) and also their attenuation due to the spreading (diffraction lobes) across a given hemisphere.
diffusers actually create "more reflections" which in turn (superposition with the direct signal) create "more comb-filters" which is overall beneficial and better "well-mixing" of the indirect sound-field over a given area in 3space. comb-filtering isn't inherently bad - just sparse comb-filtering. diffusers exist to create more and denser "comb-filters".
This means that a reduction of the reverberation time, for example, can lead to a stronger perception of the negative effects of early reflections.
a concert hall has sufficient volume (and mean free path) as to support the development of a statistically random-incidence reverberant sound-field. small acoustical spaces such as home residential rooms lack the volume to support "reverberation time". in small rooms, there exist no reverberant field that is high enough in gain to mask the early reflections (contrasted to a typical concert hall).
Also very negative are single very early (little attenuated) reflections, such as those caused by a mixing console in the studio or by the living room table at home.
and these are highly destructive due to their very-early (1-3ms) arrival time, which can have drastic image-shift effects (especially apparent with binaural source material). further compounded by the fact they are colored due to size of desk/console with respect to wavelength.
The whole context is much too complex for a simple rule like "avoid early reflections up to 15ms" or "...never feed it after midnight".
requirements and real-world loudspeaker characteristics drive the recommendations. per toole, a loudspeaker with poor off-axis response will almost always want sidewall reflections absorbed due to the spectral changes of the reflection and subsequent tonal changes imposed on perception of the direct signal. a well-behaved off-axis loudspeaker opens to the recommendation to allow small room natural sidewalls to impose sparse early first-order reflections, if subjective preferences call for such (at the expense of accuracy).