The so-called 'mass gaps' for black holes, according to theoretical models, are between 2-5 solar masses and 50 to 150 solar masses. (Actually, I have read that there is no good theoretical reason for the lower, 2 to 5 solar-mass gap....)
The lower mass-gap is suspected observationally because we have yet to observe a neutron star with mass greater than about 2 M$_{\odot}$ (the error bars on such measurements vary greatly depending on the method of observation, and its controversial currently as to what neutron star is the most massive currently known), and because we've also not discovered a black hole in an X-ray binary whose mass is lower than about 6 M$_{\odot}$. However, a binary neutron-star merger can produce a black hole which could be in this lower mass-gap. Thus, in reality, the gap will likely be populated by high-mass neutron stars (if they rotate very fast and have inclination, etc...) and by low-mass black holes that result from binary neutron star mergers.
The current nomenclature for the masses of observed black holes reflects the uncertainty inherent in the field. Right now, although the limits are arbitrary: stellar mass means $\lesssim 100$ M$_\odot$, intermediate mass means $\sim 1000 - 10^5$ M$_\odot$, supermassive means $\gtrsim 10^6$ M$_\odot$. There is no upper limit on the mass of the black hole from general relativity, but astrophysical and cosmological considerations, though model dependent, can yield an upper limit of about $10^{11}$, but it could be larger. These have been called "stupendously big" and "ultramassive."
Theoretically, the bounds on this lower mass-gap are motivated from below by stellar evolution, e.g. neutron stars are expected to only form up to a certain mass (depending on the metallicity), and astrophysical processes such as accretion. Very importantly, the bounds depend on the equation of state of nuclear material, which is uncertain and one of the "holy grails" that gravitational-wave (GW) astronomers are after.
So people often just write a range of $2 - 5$ for simplicity when writing in pop sci articles. The upper mass-gap attributed to pair-instability supernova is generally thought to be around $50 - 150$ M$_{\odot}$ (this is the black hole mass), but these limits are uncertain because they depend on the uncertainties of stellar evolution models and supernova physics. Just like with the lower mass-gap, this gap could, in practice, be populated by black holes that form from hierarchical mergers in dense stellar clusters (dynamical channel), or from isolated binary-star evolution that can in principle produce high stellar-mass binaries like GW190521 that survive failed supernova, although it is uncertain because this depends on stellar winds and core mass calculations.
EDIT: As Rob explains below, the history of stellar wind mass-loss theories is an active field. The progenitors of stellar-mass black holes are high mass stars (i.e. $m_{\rm ZAMS} \gtrsim 30$ M$_{\odot}$). For example, Wolf-Rayet stars have been viewed as the end-point of stellar nuclear evolution of high-mass stars since the 1980s. WR stars have strong winds for their size, which are line-driven, similar to the O-type stars that they likely evolve from. The mass-loss rate from the stellar surface is typically modeled as a power law of the star's luminosity and metallicity, i.e. higher mass and higher metallicity both imply more mass loss. So an initially high-mass star will experience strong winds, even more so if it has high metallicity (as Rob points out, low-metallicity environments might be loci of black holes near/in our galaxy). However, as you might hear/read in pop-sci and scientific articles, this does not preclude the existence of high-mass black holes! The most massive stars that we have observed to date are WR stars, implying they might have evolved from even more massive stars, and in the extreme case, supermassive stars are thought to be the "seeds" of supermassive black holes.
But, I have also read that astrophysicists were surprised to find black holes larger than 15 solar masses using the LIGO gravitational-wave Observatory....
I suspect you've read this from pop science articles, which often do not properly characterize the sentiment of a scientific article, but it's a difficult thing to do. The reason that anybody would be "surprised" to find black holes with masses greater than 15 M$_{\odot}$, is that before the discovery of BBHs (from the direct detections of gravitational waves) we only had suggestive evidence for the existence of stellar-mass black holes from observations of X-ray binaries. See here and here for reviews of known stellar-mass black holes in X-ray binaries.
“These observations are very hard to explain with astrophysical scenarios and they are quite easily explained with primordial black holes,” says Sébastien Clesse, a cosmologist at the University of Brussels in Belgium."
Again, this just depends on who you are asking. Everyone wants to say that their work is the most relevant formation channel for what LIGO/Virgo are seeing, and this is okay because the number of known sources is still too small to make rigorous claims about formation of BBH sources from GW detections. Also, since the signal-to-noise that the current detectors can achieve is so low (e.g. rarely above ~20), it is very hard to constrain the possible formation of even a single event. So many events that are on the edge of the current theoretical uncertainties are called "difficult to explain." This quote that you provided is perfect demonstration of this, since there actually is no evidence that priomordial black holes even exist to begin with!!!
For the event that you specifically asked about, GW190814, it is difficult mainly because of the very low mass of the secondary. But, this could be explained by, for example, isolated evolution of a binary star with an initial very low mass ratio, where the less massive component could evolve into a neutron star or a black hole. There are people who study isolated binary evolution and they claim that all LIGO sources can be explained by their channel, but they tune their simulations to produce the LIGO results. There are people who study dynamical channel evolution and claim that there is strong evidence for this or that source to be from this channel. People have to write papers and fund their students, and they do a lot of the ground-work that will make more rigorous investigations plausible once we do have larger samples of known sources. In the coming decades, it is expected that the third generation of GW detectors will provide on the order of MILLIONS of detections and then population-level constraints on formation will be much more convincing, and these detectors will have very high signal-to-noise system and constraining formation of individual events will also be more convincing.
I read elsewhere that, before LIGO, no stellar-black-holes had been detected above about 15.65 solar masses, but that reference did not say that none were expected above that mass....
I hope I've clarified this by now! If you read papers leading up to 2015 from the GW community, for example this review, they were certainly expecting to see binary black holes with masses greater than 15 M$_\odot$.